Setup docs GitHub - #3
Conversation
…kflow docs - Bump all actions to exact latest stable versions across 9 workflows: checkout@v7.0.1, codeql-action@v4.37.6, github-script@v9.0.0, osv-scanner-action@v2.5.0, release-please-action@v5.0.0, setup-python@v7.0.0, lychee-action@v2.9.0, scorecard-action@v2.4.4, action-semantic-pull-request@v6.1.1, stale@v11.0.0 - Fix repo-lint stale-reference check self-match and KhazP/Reddit-to-AI false positive - Fix labels.yml stale 'Vibe-coding' workflow label description - Add tool adapters for opencode, gemini-cli, copilot, windsurf, lovable - Add .claude skills, templates/, examples/, scripts/validate.py, CITATION.cff - Update AGENTS/MEMORY/agent_docs for Gemini 3.x model routing correction - Ignore session-ses_0*.md
- Click-by-click setup for GitHub settings, Gemini key, Cloudflare Worker secret (GEMINI_API_KEY), ALLOWED_ORIGIN var, Pages deploy, and code placeholders, so the first push runs green - Fix broken github.com/zarishsphere links in docs (org is zsdotcom) - Ignore wrangler's .dev.vars local secret file
📝 WalkthroughWalkthroughThe change adds the ZarishDocs workflow system, Claude skills, tool adapters, repository governance, CI automation, validation tooling, setup documentation, and a Reddit-to-AI example project. ChangesZarishDocs workflow foundation
Estimated code review effort: 5 (Critical) | ~120 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 65f391fc60
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| ```bash | ||
| wrangler pages project create zarishdocs # once; answer the prompts | ||
| wrangler pages deploy src # from the repo root |
There was a problem hiding this comment.
When maintainers follow this setup, Cloudflare Pages uploads only the files under src/, but the actual static entrypoint and shell assets (index.html, styles.css, sw.js, manifest.webmanifest, icon.svg) live at the repository root. Because Wrangler's positional argument is the directory of static files to upload, this deployment would publish a site with no index.html and break the app; point the command at . or at a build output that contains the root assets.
Useful? React with 👍 / 👎.
| if adapters_dir.is_dir(): | ||
| targets.extend(adapters_dir.rglob("*.md")) |
There was a problem hiding this comment.
This guard says it covers skills/tool adapters, but it only scans *.md files under templates/tool-adapters; the committed Cursor always-on rule is templates/tool-adapters/cursor/rules/zsdocs.mdc. If the forbidden cursor.sh, ai.autoComplete, or ai.explainCode strings are reintroduced there, python scripts/validate.py --ci and the Repo Lint workflow still pass, defeating the contract this check is meant to enforce.
Useful? React with 👍 / 👎.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
Note
Due to the large number of review comments, Critical severity comments were prioritized as inline comments.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
templates/agent_docs/project_brief.md (1)
10-13: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winRequire approval before payment or authentication changes.
If a project needs payment or authentication, require explicit human approval before the agent selects or modifies the integration. The Stripe Checkout example currently says to use it whenever it is simpler, so it can bypass this approval gate.
Proposed clarification
- If a simpler low-code integration exists (e.g. using pre-built Stripe Checkout instead of a custom form), use it. + If a simpler low-code integration exists, propose it. Do not select or modify payment gateway or authentication configuration without explicit human approval.Based on learnings: “Do not modify payment gateway configuration or authentication setup without explicit human approval.”
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/agent_docs/project_brief.md` around lines 10 - 13, Update the “Key Principles” guidance to require explicit human approval before selecting or modifying any payment integration, payment gateway configuration, or authentication setup. Clarify that the simpler Stripe Checkout example remains subject to this approval gate and must not be used automatically to bypass it.Source: Learnings
🟠 Major comments (36)
.github/workflows/codeql.yml-28-28 (1)
28-28: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winSet
persist-credentials: falseon both checkout steps. Both workflows check out the repository with the default credential-persistence behavior, which writes theGITHUB_TOKENinto.git/config. Neither workflow needs the token after checkout..github/workflows/scorecard.ymlline 24 already sets the flag, so the three checkout steps in this pull request are inconsistent.
.github/workflows/codeql.yml#L28-L28: addwith: persist-credentials: falseto theactions/checkout@v7.0.1step..github/workflows/repo-lint.yml#L14-L14: addwith: persist-credentials: falseto theactions/checkout@v7.0.1step. Thegit ls-filescall inscripts/validate.pydoes not need credentials, so this is safe.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/workflows/codeql.yml at line 28, Set persist-credentials to false on the checkout steps using actions/checkout@v7.0.1 in .github/workflows/codeql.yml at lines 28-28 and .github/workflows/repo-lint.yml at lines 14-14; add the corresponding with configuration to both steps, while leaving the existing checkout behavior otherwise unchanged.Source: Linters/SAST tools
.github/workflows/repo-lint.yml-1-6 (1)
1-6: 📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick winConfirm explicit approval for the nine new workflow files. This pull request adds nine GitHub Actions workflows. The pull request description states no implemented changes, related issues, or completed checklist items, so the approval is not recorded anywhere in the pull request. Record the approval in the pull request description or in
MEMORY.mdbefore merging.As per coding guidelines: "
.github/workflows/**: Do not modify GitHub workflow files without explicit approval."
.github/workflows/repo-lint.yml#L1-L6: this workflow becomes a required merge gate; confirm the approval and the intended branch-protection configuration..github/workflows/codeql.yml#L1-L5: confirm the approval for weekly code scanning on the declared languages..github/workflows/greetings.yml#L1-L7: confirm the approval for a workflow that comments on issues and pull requests..github/workflows/osv-scanner.yml#L1-L4: confirm the approval for the reusable vulnerability scan..github/workflows/pr-size-labeler.yml#L1-L5: confirm the approval for a workflow that writes labels..github/workflows/release-please.yml#L1-L8: confirm the approval for automated tagging and release publication, which usescontents: write..github/workflows/scorecard.yml#L1-L10: confirm the approval for publishing results to the public OpenSSF dataset..github/workflows/semantic-pr.yml#L1-L8: confirm the approval for enforcing Conventional Commit titles..github/workflows/stale.yml#L1-L6: confirm the approval for automatic closure of issues and pull requests.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/workflows/repo-lint.yml around lines 1 - 6, Record explicit approval for all nine workflow files in the pull request description or MEMORY.md before merging: .github/workflows/repo-lint.yml (1-6), .github/workflows/codeql.yml (1-5), .github/workflows/greetings.yml (1-7), .github/workflows/osv-scanner.yml (1-4), .github/workflows/pr-size-labeler.yml (1-5), .github/workflows/release-please.yml (1-8), .github/workflows/scorecard.yml (1-10), .github/workflows/semantic-pr.yml (1-8), and .github/workflows/stale.yml (1-6). Confirm the intended branch-protection configuration for repo-lint and approval for each workflow’s described behavior; no direct workflow-file change is required.Source: Coding guidelines
scripts/validate.py-325-329 (1)
325-329: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winThe fallback
yearregex misses a top-levelyear:key.Line 327 uses
^\s+year\s*:. The\s+requires at least one whitespace character after the line start, so ayear:key at column 0 does not match. The result also depends on whether a blank line precedes the key, because\s+can consume a blank line plus its newline.The PyYAML path at line 319 accepts a top-level
year. The two paths therefore disagree on the same file.
.github/workflows/repo-lint.ymllines 85-93 sets up Python and runs the validator without installing PyYAML, so CI runs this fallback path. ACITATION.cffwith a top-levelyearfails CI and passes locally for a contributor who has PyYAML installed.Match
yearat any indentation, including none.🔧 Proposed fix
- has_year = re.search(r"^\s+year\s*:", text, re.MULTILINE) is not None + has_year = re.search(r"^[ \t]*year\s*:", text, re.MULTILINE) is not None🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@scripts/validate.py` around lines 325 - 329, Update the fallback year detection in the validation logic around top_level_keys and has_year so the regex matches a year: key at any indentation level, including column zero, without allowing preceding blank lines to affect the match. Keep the PyYAML and fallback paths consistent for top-level year keys..github/workflows/pr-size-labeler.yml-53-72 (1)
53-72: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick win
setLabelscan delete labels that this workflow did not manage.Two problems combine here.
listLabelsOnIssueat line 53 is not paginated. It returns at most 30 labels.github.paginateis already used at line 24 for the file list, so the inconsistency is unintentional.
setLabelsat line 67 replaces the complete label set withnonSizeLabels + sizeLabel. Any label past the first page is deleted. Any label added by another actor between the read at line 53 and the write at line 67 is also deleted. That is a time-of-check to time-of-use race.Use
removeLabelfor stalesize/labels andaddLabelsfor the new one. Both operations are non-destructive and idempotent, and neither needs the full label list.🔧 Proposed fix
const current = await github.rest.issues.listLabelsOnIssue({ owner, repo, - issue_number: pull_number + issue_number: pull_number, + per_page: 100 }); const existingSizeLabels = current.data .map((l) => l.name) .filter((name) => name.startsWith('size/')); - const nonSizeLabels = current.data - .map((l) => l.name) - .filter((name) => !name.startsWith('size/')); - - await github.rest.issues.setLabels({ - owner, - repo, - issue_number: pull_number, - labels: [...nonSizeLabels, sizeLabel] - }); + for (const name of existingSizeLabels) { + if (name === sizeLabel) continue; + await github.rest.issues.removeLabel({ + owner, repo, issue_number: pull_number, name + }); + } + + if (!existingSizeLabels.includes(sizeLabel)) { + await github.rest.issues.addLabels({ + owner, repo, issue_number: pull_number, labels: [sizeLabel] + }); + }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/workflows/pr-size-labeler.yml around lines 53 - 72, Update the label-management flow around listLabelsOnIssue to avoid replacing the complete label set: remove each existing stale size/ label with removeLabel, then apply sizeLabel using addLabels. Use the retrieved labels only to identify stale size/ entries; preserve all unrelated labels and avoid setLabels, pagination, and read/write race issues..github/workflows/release-please.yml-18-21 (1)
18-21: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winThe release pull request will not run any status checks.
No
token:input is set, so the action uses the defaultGITHUB_TOKEN. GitHub does not trigger workflow runs for events raised by that token. The standing release pull request will therefore skiprepo-lint.yml,semantic-pr.yml,codeql.yml, andosv-scanner.yml.
.github/PULL_REQUEST_TEMPLATE.mdrequirespython scripts/validate.pyto pass on every pull request. The release pull request would merge without that gate.Supply a GitHub App token or a personal access token through the
token:input so the release pull request runs the same checks.🔧 Proposed fix
- uses: googleapis/release-please-action@v5.0.0 with: + # A GITHUB_TOKEN-created PR does not trigger workflows; use an App/PAT token + # so the release PR runs repo-lint, semantic-pr, CodeQL, and OSV. + token: ${{ secrets.RELEASE_PLEASE_TOKEN }} # Node release type bumps package.json version (v0.1.0 at the repo root). release-type: node🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/workflows/release-please.yml around lines 18 - 21, Update the release-please action configuration to provide a non-default GitHub App or personal access token through its token input, sourced from the repository’s configured secret. Keep the existing release-type: node setting unchanged so release pull requests trigger the standard status-check workflows.examples/reddit-to-ai/research-reddit-to-ai.md-5-5 (1)
5-5: 📐 Maintainability & Code Quality | 🟠 Major | 🏗️ Heavy liftAdd sources and evidence status for factual claims.
Only a few claims use
Unverified — model knowledge; most claims lack a live source and access date. Add dated sources for claims about context windows,chrome.storage, competitor behavior, costs, and timelines. Label claims that cannot be sourced asUnverified — model knowledge.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/reddit-to-ai/research-reddit-to-ai.md` at line 5, Update the research document’s factual claims to include dated sources and access dates, especially claims about context windows, chrome.storage, competitor behavior, costs, and timelines. Mark any claims that cannot be supported with a source as “Unverified — model knowledge,” while preserving the existing evidence-status format.Source: Coding guidelines
examples/reddit-to-ai/TechDesign-reddit-to-ai-MVP.md-53-54 (1)
53-54: 🎯 Functional Correctness | 🟠 Major | 🏗️ Heavy liftRun
redditScraper.jsoutside the service worker.The service worker cannot access the Reddit page DOM. Keep
service_worker.jsas the orchestrator, and runredditScraper.jsas a content script or throughchrome.scripting.executeScript; return normalized thread data through messaging. Declare the required injection and host permissions. Source: https://developer.chrome.com/docs/extensions/develop/concepts/content-scripts (accessed August 2026).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/reddit-to-ai/TechDesign-reddit-to-ai-MVP.md` around lines 53 - 54, Update the Reddit scraping flow so redditScraper.js executes in the Reddit page context as a content script or via chrome.scripting.executeScript, while service_worker.js remains the orchestrator. Return normalized thread data through messaging, and declare the required script-injection and Reddit host permissions in the extension manifest.agent_docs/tech_stack.md-24-34 (1)
24-34: 🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy liftRequire authentication or rate limiting for production proxy requests
worker/index.jsaccepts a validPOSTwithoutOriginand forwards it withenv.GEMINI_API_KEY. CORS validation does not authenticate non-browser callers. Add production authentication or rate limiting. If test calls need no authentication, gate them behind an explicit test-only environment flag.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@agent_docs/tech_stack.md` around lines 24 - 34, Update the worker fetch handler described in the Worker Proxy Sketch to require production authentication or rate limiting before forwarding valid POST requests with env.GEMINI_API_KEY. Do not treat Origin/CORS validation as authentication; if tests must bypass this check, require an explicit test-only environment flag and keep the bypass disabled by default.agent_docs/code_patterns.md-33-37 (1)
33-37: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winImplement the documented IndexedDB blocked-request handling.
During a version upgrade, another open tab can keep
indexedDB.open()pending, which blocks session persistence. Addonblockedhandling and close existing connections onversionchange, or provide an explicit retry/reload path. MDNblockedevent (accessed August 2026).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@agent_docs/code_patterns.md` around lines 33 - 37, Update the IndexedDB implementation in src/db.js, especially openDB, to handle blocked version upgrades by closing existing database connections when they receive a versionchange event and providing an explicit retry or reload path for the blocked request. Preserve the documented database stores and existing persistence API behavior.agent_docs/testing.md-5-5 (1)
5-5: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winMake
npm run checkvalidate the documented files.
package.jsondefinescheckas a loop oversrc/andworker/. It does not runnode --check sw.js, although this line claims that it does. Extend the script to includesw.js, or removesw.jsfrom the documentation.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@agent_docs/testing.md` at line 5, Align the documented Syntax check with the npm check script in package.json: update the check command to run node --check against sw.js in addition to src/ and worker/, preserving validation of all documented shipped modules.MEMORY.md-19-19 (1)
19-19: 🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy liftKeep the final File System Access API decision synchronized.
The project-state record and generated inventory still contain the obsolete file-writing architecture.
MEMORY.md#L19-L19: mark the older ADR-003 entry as superseded or update it toshowDirectoryPicker.REPOSITORY-INVENTORY.md#L100-L105: regenerate the file description and related inventory metadata from the final implementation.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@MEMORY.md` at line 19, Update MEMORY.md at line 19 to mark the obsolete ADR-003 file-writing architecture as superseded or describe the final showDirectoryPicker implementation. Regenerate REPOSITORY-INVENTORY.md lines 100-105 so the file description and related metadata match the final implementation; apply changes in both specified files.CITATION.cff-9-9 (1)
9-9: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winAlign the license metadata and ship the license texts.
CITATION.cffdeclares MIT, but project documentation declares Apache 2.0 for code and CC BY 4.0 for documentation. The repository tracks no license file. Add the corresponding license texts and update the metadata, or correct the documentation and metadata to match the supported licenses.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CITATION.cff` at line 9, Align CITATION.cff with the project’s documented licenses by using Apache 2.0 for code and CC BY 4.0 for documentation, and add the corresponding license text files to the repository. Remove the conflicting MIT declaration from the license metadata while preserving valid citation formatting..github/CODEOWNERS-5-7 (1)
5-7: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winRequire a visible team with explicit Write access.
GitHub requires teams used as code owners to be visible and have explicit
Writeaccess. ReplaceReadand confirm thatzsdotcom/maintainersexists and has access tozsdotcom/zarishdocs; the API lookup currently returns404 Not Found.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/CODEOWNERS around lines 5 - 7, Update the CODEOWNERS prerequisite comment to require referenced teams to be visible and have explicit Write access instead of Read access. Also verify that the zsdotcom/maintainers team exists and has access to zsdotcom/zarishdocs, correcting the team or repository configuration if the lookup remains 404.SETUP.md-29-32 (1)
29-32: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winSeparate repository-visibility requirements by tool.
- CodeQL is free for public repositories. Private repositories require GitHub Code Security or GHAS (GitHub documentation, accessed 2026-08-11).
- OSV-Scanner runs in private repositories and does not require public visibility (OSV-Scanner documentation, accessed 2026-08-11).
- Scorecard can scan private repositories. Only
publish_resultsrequires a public repository (Scorecard Action documentation, accessed 2026-08-11).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@SETUP.md` around lines 29 - 32, Update the “Confirm repo visibility” section in SETUP.md to distinguish requirements by tool: state that CodeQL’s free availability applies to public repositories, OSV-Scanner supports private repositories, and Scorecard supports private scanning while its publish_results option requires public visibility. Remove the blanket requirement that the repository must be public.Source: Coding guidelines
CITATION.cff-16-18 (1)
16-18: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winMake
CITATION.cffrelease metadata valid and consistent.
date-releasedis optional, but an empty string is invalid. If the project is unreleased, omit this field. Otherwise, set it to the actualYYYY-MM-DDrelease date. Alignpreferred-citation.versionwith the0.1.0version declared inpackage.jsonandSETUP.md.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CITATION.cff` around lines 16 - 18, Update the release metadata in CITATION.cff: remove the empty date-released field if the project is unreleased, or replace it with the actual YYYY-MM-DD release date, and change preferred-citation.version from 1.0.0 to 0.1.0 to match package.json and SETUP.md.SETUP.md-192-198 (1)
192-198: 🩺 Stability & Availability | 🟠 Major | 🏗️ Heavy liftDeploy a complete Pages asset directory.
wrangler pages deploy srcomits the root-level entrypoint and assets referenced byindex.html, includingstyles.css,manifest.webmanifest,icon.svg,sw.js, andsources.config.json. Stage these files withsrc/, then deploy the staging directory.Proposed staging command
rm -rf dist mkdir -p dist cp index.html styles.css sw.js manifest.webmanifest icon.svg sources.config.json dist/ cp -R src dist/src wrangler pages deploy distSource: Cloudflare Pages Direct Upload (accessed August 2026).
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@SETUP.md` around lines 192 - 198, Update the Pages deployment instructions in “Create and deploy the Pages site” to stage the complete asset set in a clean dist directory, including index.html, styles.css, sw.js, manifest.webmanifest, icon.svg, sources.config.json, and src/, then deploy dist instead of src..github/ISSUE_TEMPLATE/config.yml-3-5 (1)
3-5: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winExpose the fallback security contact in the issue chooser.
If private vulnerability reporting is unavailable, direct reporters to
platform@zarishsphere.com, the fallback address in.github/SECURITY.md, while retaining the instruction not to open a public issue.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/ISSUE_TEMPLATE/config.yml around lines 3 - 5, Update the “Security vulnerability” entry in the issue template chooser to expose platform@zarishsphere.com as the fallback contact when private reporting is unavailable, while retaining the instruction not to open a public issue..github/MCP-SETUP.md-1-8 (1)
1-8: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winSeparate local and remote server setup, and qualify the cost statement.
The Remote option uses GitHub’s hosted MCP server. The Docker option runs locally. Replace the opening hosting claim with this distinction. Replace
Freewith a statement that the server is available to GitHub users, while individual tools can require paid GitHub or Copilot features and organization policies.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/MCP-SETUP.md around lines 1 - 8, Update the opening description in the MCP setup documentation to distinguish GitHub’s hosted Remote MCP server from the locally running Docker option. Replace the unconditional “Free” claim with qualified wording that availability is for GitHub users, while individual tools may require paid GitHub or Copilot features and remain subject to organization policies..github/SECURITY.md-16-18 (1)
16-18: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winLimit the policy scope to this repository.
A repository-local
.github/SECURITY.mdtakes precedence over any organization default. Change line 18 to:This applies to the `zsdotcom/zarishdocs` repository.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/SECURITY.md around lines 16 - 18, Update the Scope statement in SECURITY.md to limit the policy to this repository by replacing the organization-wide scope with “This applies to the `zsdotcom/zarishdocs` repository.”templates/tool-adapters/lovable/README.md-7-12 (1)
7-12: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winFix the no-repository Knowledge instructions.
If no repository is connected,
AGENTS.mdandagent_docs/are unavailable. The quoted Project Knowledge pointer therefore gives Lovable no project rules. Put the required rules directly in Project Knowledge, or require a connected repository before telling users to use the pointer.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/lovable/README.md` around lines 7 - 12, Update the no-repository guidance in the README so it does not direct users to the unavailable AGENTS.md or agent_docs/ pointer; provide the required rules directly in Project Knowledge, or require connecting a repository before recommending that pointer.templates/tool-adapters/cursor/commands/zsdocs-workflow.md-18-18 (1)
18-18: 🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy liftDefine one required
Handoff Contextschema across all adapters.The research and PRD adapters define only
Stage, and the tech-design adapter adds onlyChosen stackandAI coding tool. This does not satisfy the router contract for user level, app name, platform, budget, and timeline. Update every adapter variant, or narrow the router statement to the fields each stage guarantees.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/cursor/commands/zsdocs-workflow.md` at line 18, Align the router contract and adapter documentation around a single required Handoff Context schema: ensure every research, PRD, and tech-design adapter defines user level, app name, platform, budget, and timeline alongside its existing fields. Alternatively, narrow the router statement to explicitly list only the fields guaranteed by each stage, without claiming universal coverage.templates/tool-adapters/README.md-57-59 (1)
57-59: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winPreserve existing agent rules unless the user confirms deletion.
The same unsafe deletion instruction appears in three adapters.
templates/tool-adapters/README.md#L57-L59: replace “delete it” with an explicit confirmation step.templates/tool-adapters/windsurf/README.md#L11-L13: preserve.windsurfrulesby default..claude/skills/zsdocs-agents/SKILL.md#L170-L172: do not delete.cursorrulesautomatically.As per coding guidelines, “Do not delete files without explicit confirmation.”
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/README.md` around lines 57 - 59, Update the legacy rule guidance to preserve existing files unless the user explicitly confirms deletion: in templates/tool-adapters/README.md lines 57-59, replace the “delete it” instruction with a confirmation step; in templates/tool-adapters/windsurf/README.md lines 11-13, state that .windsurfrules is preserved by default; and in .claude/skills/zsdocs-agents/SKILL.md lines 170-172, ensure .cursorrules is not deleted automatically.Source: Coding guidelines
.claude/skills/zsdocs-agents/SKILL.md-75-86 (1)
75-86: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winRequire an explicit plan and approval before agent setup writes multiple files.
Both agent-setup entry points can generate a large file set without a required approval checkpoint.
.claude/skills/zsdocs-agents/SKILL.md#L75-L86: add a plan and wait for explicit confirmation before instantiating templates.templates/tool-adapters/antigravity/workflows/zsdocs-agents.md#L9-L16: add the same approval gate before generating project files.Based on learnings, “Before changing more than one file, propose a brief step-by-step plan and wait for approval.”
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-agents/SKILL.md around lines 75 - 86, Require an explicit plan and approval before generating project files in both entry points: .claude/skills/zsdocs-agents/SKILL.md lines 75-86 and templates/tool-adapters/antigravity/workflows/zsdocs-agents.md lines 9-16. Update the template-instantiation workflow to propose a brief step-by-step plan, wait for explicit confirmation, and only then copy the multiple template files; apply the same approval gate in the Antigravity workflow.Source: Learnings
.claude/skills/zsdocs-agents/SKILL.md-15-19 (1)
15-19: 🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy liftKeep imported project text out of persistent agent instructions.
Lines 15-19 require continuity hints, and Lines 88-104 copy PRD and Tech Design content into
AGENTS.md,MEMORY.md, andagent_docs. If those documents contain imported or untrusted instruction text, it can become persistent prompt injection for future sessions. Extract only allowlisted fields, delimit imported prose as project data, and show the generated-file diff for approval before writing.The supplied persistent-context warning is consistent with this data flow.
Also applies to: 88-104
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-agents/SKILL.md around lines 15 - 19, Update the Session Continuity guidance and the generation flow around the referenced AGENTS.md, MEMORY.md, and agent_docs outputs to prevent imported PRD or Tech Design prose from becoming persistent agent instructions. Extract only explicitly allowlisted fields, clearly delimit retained project text as untrusted project data, and require displaying the generated-file diff for user approval before writing any files.Source: Linters/SAST tools
.claude/settings.json-19-19 (1)
19-19: 🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy liftClose bypasses in the destructive-command guard.
The
rmexpression matches only literal target forms. It does not match equivalent commands such asrm -rf "$PWD",rm -rf /home/user/project, orrm -rf ./*. TheTRUNCATEexpression is case-sensitive, so lowercasetruncate tablebypasses it. Implement a path-aware command policy or narrow the README claim, then test quoted paths, command substitutions, long flags, and case variants.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/settings.json at line 19, Strengthen the destructive-command guard in the inline Node command by detecting equivalent recursive force-removal targets, including quoted paths, $PWD-based paths, absolute project paths, wildcard forms, and long flags; make the TRUNCATE check case-insensitive. Add coverage for quoted paths, command substitutions, long flags, and case variants, or narrow the README claim if the guard cannot provide that coverage.templates/tool-adapters/opencode/commands/zsdocs-workflow.md-9-20 (1)
9-20: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winDo not route incomplete agent setup to build.
Lines 9-10 only check
AGENTS.mdand theagent_docs/directory. The setup contract requiresMEMORY.md,REVIEW-CHECKLIST.md, and all fiveagent_docs/*.mdfiles. A directory can exist while those files are absent, so Line 20 can route an incomplete project to/zsdocs-build.Check every required output or use a generated completion marker.
Proposed routing check
-1. Check which of these exist: `docs/research-*.md`, `docs/PRD-*.md`, `docs/TechDesign-*.md`, `AGENTS.md` + `agent_docs/`. +1. Check which of these exist: `docs/research-*.md`, `docs/PRD-*.md`, `docs/TechDesign-*.md`, `AGENTS.md`, `MEMORY.md`, `REVIEW-CHECKLIST.md`, and all required `agent_docs/*.md`.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/opencode/commands/zsdocs-workflow.md` around lines 9 - 20, Update the stage-detection logic in the zsdocs workflow to verify complete agent setup before selecting /zsdocs-build: require AGENTS.md, MEMORY.md, REVIEW-CHECKLIST.md, and all five required agent_docs/*.md files, or use the established generated completion marker. Route to /zsdocs-agents whenever any required setup output is missing, rather than treating the directory’s existence as completion..claude/settings.json-3-9 (1)
3-9: 🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy liftEnforce protected paths for Bash, or document the limited scope.
The file-protection hook matches only
Edit|Write. The separateBashhook does not check protected paths, so shell redirection or copy commands can modify.env, lockfiles, and.gitinternals. The check also uses unresolved lexical paths, so symlink aliases are not protected. Update the policy to resolve paths, or state theEdit|Writelimitation in.claude/README.md.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/settings.json around lines 3 - 9, Extend the protected-path enforcement in the PreToolUse configuration to cover Bash-based file modifications, including redirection and copy commands, and resolve canonical paths before checking .env files, lockfiles, and .git internals so symlink aliases are blocked. Alternatively, document the existing Edit|Write-only limitation in the Claude policy README, but ensure the chosen approach clearly defines the protection scope..claude/skills/zsdocs-build/SKILL.md-82-102 (1)
82-102: 🔒 Security & Privacy | 🟠 Major | ⚡ Quick winMake infrastructure and analytics steps conditional.
Database setup, authentication, and analytics are written as mandatory phases. Execute them only when the approved PRD and Tech Design require them. Otherwise, this skill can add infrastructure or tracking outside the project scope.
As per coding guidelines, do not add features outside the current phase.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-build/SKILL.md around lines 82 - 102, Update the phase guidance so database connection, authentication, and analytics are conditional on explicit requirements in the approved PRD and Tech Design, rather than mandatory steps. State that infrastructure and tracking must not be added when unspecified, and preserve the rule against introducing features outside the current phase.Source: Coding guidelines
.claude/skills/zsdocs-prd/SKILL.md-46-101 (1)
46-101: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winMake every required handoff field collectable before generating the next artifact.
Later stages require fields that the question banks do not guarantee.
.claude/skills/zsdocs-prd/SKILL.md#L46-L101: collect and confirm platform, budget, and timeline for every fresh PRD path..claude/skills/zsdocs-techdesign/SKILL.md#L50-L73: ask which AI coding tool the user will use..claude/skills/zsdocs-techdesign/SKILL.md#L98-L114: do not requireAI coding toolin Handoff Context unless it was collected.templates/tool-adapters/antigravity/workflows/zsdocs-techdesign.md#L10-L14: mirror the same required-field collection in the adapter.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-prd/SKILL.md around lines 46 - 101, Update .claude/skills/zsdocs-prd/SKILL.md lines 46-101 so every fresh PRD path collects and confirms platform, budget, and timeline before generating the next artifact, including Level B and Level C where currently missing. Update .claude/skills/zsdocs-techdesign/SKILL.md lines 50-73 to ask which AI coding tool the user will use, and update lines 98-114 so AI coding tool is included in Handoff Context only when collected. Mirror the same required-field collection in templates/tool-adapters/antigravity/workflows/zsdocs-techdesign.md lines 10-14..claude/skills/zsdocs-workflow/SKILL.md-90-128 (1)
90-128: 🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy liftEnforce all stage prerequisites before advancing.
The
If Tech Design Existsbranch can start Agent Setup without checking the PRD. TheIf AGENTS.md Existsbranch marks the project ready to build without checkingagent_docs/, PRD, or Tech Design. The build skill requires all four files.Use one first-missing-stage decision instead of existence-only branches.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-workflow/SKILL.md around lines 90 - 128, Update the stage-routing logic in the “If Tech Design Exists” and “If AGENTS.md Exists” sections to use a single first-missing-stage decision that enforces the required sequence: Research, PRD, Tech Design, then Agent Config with AGENTS.md, MEMORY.md, REVIEW-CHECKLIST.md, and agent_docs/. Only announce Agent Config when all earlier artifacts exist, and only start the build when every required artifact is present; otherwise route the user to the earliest missing stage..claude/skills/zsdocs-prd/SKILL.md-33-44 (1)
33-44: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winConsume
Handoff Contextbefore repeating stage questions.The canonical skills and Antigravity adapter do not consistently use persisted handoff data.
.claude/skills/zsdocs-prd/SKILL.md#L33-L44: skip the A/B/C question when the research handoff is valid..claude/skills/zsdocs-techdesign/SKILL.md#L40-L48: skip the A/B/C question when the PRD handoff is valid.templates/tool-adapters/antigravity/workflows/zsdocs-techdesign.md#L10-L14: read the PRD handoff before asking the technical-level question.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-prd/SKILL.md around lines 33 - 44, Update .claude/skills/zsdocs-prd/SKILL.md lines 33-44, .claude/skills/zsdocs-techdesign/SKILL.md lines 40-48, and templates/tool-adapters/antigravity/workflows/zsdocs-techdesign.md lines 10-14 to read and validate the persisted Handoff Context before asking technical-level questions; when valid, prefill and confirm its values and skip the A/B/C prompt, otherwise retain the existing question flow.templates/tool-adapters/codex/prompts/zsdocs-agents.md-10-16 (1)
10-16: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winDo not derive every AGENTS value only from PRD and Tech Design.
The canonical
templates/AGENTS.mdalso requires setup, test, lint, build, formatter, architecture, and current-state values. If a value is absent, inspect the existing project/toolchain or ask the user. Do not invent package-manager commands.The canonical
templates/AGENTS.mdcontract supplies project-specific commands and forbids invented package-manager commands.Proposed prompt change
-- fill every `[REPLACE:]` / `[CHOOSE:]` placeholder from the PRD and Tech Design; never rewrite the templates: +- fill placeholders from the PRD and Tech Design; inspect the project or ask the user for missing values; never invent commands or rewrite the templates:🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/codex/prompts/zsdocs-agents.md` around lines 10 - 16, Update the canonical-template instantiation guidance in the prompt to require deriving AGENTS.md setup, test, lint, build, formatter, architecture, and current-state values from the existing project/toolchain or the user when absent from the PRD and Tech Design. Explicitly prohibit inventing package-manager commands while preserving the requirement to fill every template placeholder..claude/skills/zsdocs-build/SKILL.md-68-76 (1)
68-76: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winUse project-defined verification commands.
Hardcoded
npm testandnpm run lintare invalid for non-Node projects and conflict withAGENTS.md, which stores the project commands.Proposed verification change
-1. Run tests: `npm test` (or equivalent) -2. Run linter: `npm run lint` +1. Run the test command defined in `AGENTS.md` or `agent_docs/tech_stack.md` +2. Run the lint and formatting commands defined in those filesThe canonical AGENTS contract defines project-specific commands and forbids invented package-manager commands.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-build/SKILL.md around lines 68 - 76, Update the verification steps in the feature workflow to use the project-defined test, lint, and other verification commands documented in AGENTS.md instead of hardcoded npm commands. Preserve the required smoke testing, issue fixing, and AGENTS.md/MEMORY.md updates, and retain browser verification for frontend projects..claude/skills/zsdocs-build/SKILL.md-4-4 (1)
4-4: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winMake verification and build steps project-specific.
- Add a supported browser/computer tool, or require the user to run the required Chromium and Safari/Firefox/mobile checks before completion.
- Replace the hardcoded
npm run lintstep with the project-defined checks, includingnpm test,npm run check, andnpm run format:check.- Remove mandatory database, authentication, and analytics steps when they are not in
AGENTS.md; analytics conflicts with the repository’s no-telemetry constraint.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/zsdocs-build/SKILL.md at line 4, Update the verification and build instructions in SKILL.md to use project-defined checks, including npm test, npm run check, and npm run format:check instead of hardcoded linting. Add a supported browser/computer tool or require the user to perform Chromium and Safari/Firefox/mobile checks before completion, and remove database, authentication, and analytics requirements unless AGENTS.md explicitly requires them.Source: Coding guidelines
templates/tool-adapters/codex/prompts/zsdocs-workflow.md-9-20 (1)
9-20: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winCheck all Stage 4 artifacts before routing to Build.
Line [9] treats
AGENTS.mdplusagent_docs/as complete Stage 4 setup.templates/tool-adapters/cursor/commands/zsdocs-agents.mdalso createsMEMORY.mdandREVIEW-CHECKLIST.mdat Lines [8]-[12]. The build adapters requireREVIEW-CHECKLIST.md. If it is missing, this router can skip setup and send the user to Build without a required checklist. Check all Stage 4 artifacts before selecting stage 5.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/codex/prompts/zsdocs-workflow.md` around lines 9 - 20, Update the Stage 4 artifact check in the workflow routing instructions to require AGENTS.md, agent_docs/, MEMORY.md, and REVIEW-CHECKLIST.md before selecting Stage 5. Keep routing to /zsdocs-agents when any required setup artifact is missing, while preserving the existing stage ordering and behavior for earlier stages.templates/tool-adapters/codex/prompts/zsdocs-workflow.md-5-22 (1)
5-22: 🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy liftKeep all Codex and Cursor adapters aligned with the canonical workflow contract.
templates/tool-adapters/README.mddefineszsdocs-*files as thin wrappers around rootpart*.mdfiles. These eight files embed complete workflows, so canonical changes can drift across tools.
templates/tool-adapters/codex/prompts/zsdocs-workflow.md#L5-L22: reduce the router to a thin Codex wrapper.templates/tool-adapters/codex/prompts/zsdocs-build.md#L5-L15: reduce the build command to a thin Codex wrapper.templates/tool-adapters/codex/prompts/zsdocs-prd.md#L5-L17: reduce the PRD command to a thin Codex wrapper.templates/tool-adapters/codex/prompts/zsdocs-research.md#L5-L17: reduce the research command to a thin Codex wrapper.templates/tool-adapters/codex/prompts/zsdocs-techdesign.md#L5-L17: reduce the technical-design command to a thin Codex wrapper.templates/tool-adapters/cursor/commands/zsdocs-agents.md#L1-L14: reduce the agent-setup command to a thin Cursor wrapper.templates/tool-adapters/cursor/commands/zsdocs-build.md#L1-L11: reduce the build command to a thin Cursor wrapper.templates/tool-adapters/cursor/commands/zsdocs-prd.md#L1-L13: reduce the PRD command to a thin Cursor wrapper.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@templates/tool-adapters/codex/prompts/zsdocs-workflow.md` around lines 5 - 22, Reduce the complete workflow content in templates/tool-adapters/codex/prompts/zsdocs-workflow.md (lines 5-22), zsdocs-build.md (lines 5-15), zsdocs-prd.md (lines 5-17), zsdocs-research.md (lines 5-17), zsdocs-techdesign.md (lines 5-17), templates/tool-adapters/cursor/commands/zsdocs-agents.md (lines 1-14), zsdocs-build.md (lines 1-11), and zsdocs-prd.md (lines 1-13) to thin wrappers that delegate to the corresponding canonical root part*.md workflow, following the contract defined in templates/tool-adapters/README.md; remove duplicated workflow instructions while preserving each adapter’s command-specific delegation.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI (base), Organization UI (inherited)
Review profile: CHILL
Plan: Pro Plus
Run ID: 68a3099e-9c89-4985-a2a4-39e60e523467
📒 Files selected for processing (99)
.claude/README.md.claude/settings.json.claude/skills/zsdocs-agents/SKILL.md.claude/skills/zsdocs-build/SKILL.md.claude/skills/zsdocs-prd/SKILL.md.claude/skills/zsdocs-research/SKILL.md.claude/skills/zsdocs-techdesign/SKILL.md.claude/skills/zsdocs-workflow/SKILL.md.github/CODEOWNERS.github/CODE_OF_CONDUCT.md.github/CONTRIBUTING.md.github/DISCUSSION_TEMPLATE/governance.yml.github/FAQ.md.github/FUNDING.yml.github/GOVERNANCE.md.github/ISSUE_TEMPLATE/bug_report.yml.github/ISSUE_TEMPLATE/config.yml.github/ISSUE_TEMPLATE/feature_request.yml.github/ISSUE_TEMPLATE/showcase_submission.md.github/MCP-SETUP.md.github/PULL_REQUEST_TEMPLATE.md.github/SECURITY.md.github/SUPPORT.md.github/labels.yml.github/workflows/codeql.yml.github/workflows/greetings.yml.github/workflows/osv-scanner.yml.github/workflows/pr-size-labeler.yml.github/workflows/release-please.yml.github/workflows/repo-lint.yml.github/workflows/scorecard.yml.github/workflows/semantic-pr.yml.github/workflows/stale.yml.gitignoreAGENTS.mdCITATION.cffMEMORY.mdREADME.mdREPOSITORY-INVENTORY.mdSETUP.mdagent_docs/code_patterns.mdagent_docs/product_requirements.mdagent_docs/project_brief.mdagent_docs/tech_stack.mdagent_docs/testing.mddocs/TechDesign-ZarishDocs-MVP.mddocs/research-ZarishDocs.mdexamples/README.mdexamples/reddit-to-ai/AGENTS.mdexamples/reddit-to-ai/MEMORY.mdexamples/reddit-to-ai/PRD-reddit-to-ai-MVP.mdexamples/reddit-to-ai/TechDesign-reddit-to-ai-MVP.mdexamples/reddit-to-ai/research-reddit-to-ai.mdscripts/validate.pytemplates/AGENTS.mdtemplates/MEMORY.mdtemplates/REVIEW-CHECKLIST.mdtemplates/agent_docs/code_patterns.mdtemplates/agent_docs/product_requirements.mdtemplates/agent_docs/project_brief.mdtemplates/agent_docs/tech_stack.mdtemplates/agent_docs/testing.mdtemplates/tool-adapters/CLAUDE.mdtemplates/tool-adapters/README.mdtemplates/tool-adapters/antigravity/rules/zsdocs.mdtemplates/tool-adapters/antigravity/workflows/zsdocs-agents.mdtemplates/tool-adapters/antigravity/workflows/zsdocs-build.mdtemplates/tool-adapters/antigravity/workflows/zsdocs-prd.mdtemplates/tool-adapters/antigravity/workflows/zsdocs-research.mdtemplates/tool-adapters/antigravity/workflows/zsdocs-techdesign.mdtemplates/tool-adapters/antigravity/workflows/zsdocs-workflow.mdtemplates/tool-adapters/codex/README.mdtemplates/tool-adapters/codex/prompts/zsdocs-agents.mdtemplates/tool-adapters/codex/prompts/zsdocs-build.mdtemplates/tool-adapters/codex/prompts/zsdocs-prd.mdtemplates/tool-adapters/codex/prompts/zsdocs-research.mdtemplates/tool-adapters/codex/prompts/zsdocs-techdesign.mdtemplates/tool-adapters/codex/prompts/zsdocs-workflow.mdtemplates/tool-adapters/copilot/README.mdtemplates/tool-adapters/copilot/copilot-instructions.mdtemplates/tool-adapters/cursor/commands/zsdocs-agents.mdtemplates/tool-adapters/cursor/commands/zsdocs-build.mdtemplates/tool-adapters/cursor/commands/zsdocs-prd.mdtemplates/tool-adapters/cursor/commands/zsdocs-research.mdtemplates/tool-adapters/cursor/commands/zsdocs-techdesign.mdtemplates/tool-adapters/cursor/commands/zsdocs-workflow.mdtemplates/tool-adapters/cursor/rules/zsdocs.mdctemplates/tool-adapters/gemini-cli/GEMINI.mdtemplates/tool-adapters/gemini-cli/README.mdtemplates/tool-adapters/lovable/README.mdtemplates/tool-adapters/opencode/README.mdtemplates/tool-adapters/opencode/commands/zsdocs-agents.mdtemplates/tool-adapters/opencode/commands/zsdocs-build.mdtemplates/tool-adapters/opencode/commands/zsdocs-prd.mdtemplates/tool-adapters/opencode/commands/zsdocs-research.mdtemplates/tool-adapters/opencode/commands/zsdocs-techdesign.mdtemplates/tool-adapters/opencode/commands/zsdocs-workflow.mdtemplates/tool-adapters/windsurf/README.mdtemplates/tool-adapters/windsurf/rules/zsdocs.md
| "hooks": [ | ||
| { | ||
| "type": "command", | ||
| "command": "node -e \"const fs=require('fs'); const d=JSON.parse(fs.readFileSync(0,'utf8')); const p=d.tool_input?.file_path||''; if(/\\.(ts|tsx|js|jsx)$/.test(p) && fs.existsSync('node_modules/.bin/prettier')) { require('child_process').execSync('npx prettier --write \\\"'+p+'\\\"', {stdio:'inherit'}); }\"", |
There was a problem hiding this comment.
🔒 Security & Privacy | 🔴 Critical | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -eu
printf '%s\n' '--- .claude/settings.json lines 1-45 ---'
nl -ba .claude/settings.json | sed -n '1,45p'
printf '%s\n' '--- child-process invocation context ---'
rg -n -C 3 'execSync|execFileSync|prettier|tool_input|PostToolUse' .claude/settings.json
printf '%s\n' '--- path-related repository context ---'
rg -n -C 2 'settings\.json|PostToolUse|prettier --write|node_modules/\.bin/prettier' . --glob '!node_modules' --glob '!dist' --glob '!build'Repository: zsdotcom/zarishdocs
Length of output: 7991
🏁 Script executed:
#!/bin/bash
set -eu
python3 - <<'PY'
import json
from pathlib import Path
path = Path(".claude/settings.json")
data = json.loads(path.read_text())
hooks = data["hooks"]["PostToolUse"]
command = hooks[0]["hooks"][0]["command"]
print("json_parse=ok")
print("posttooluse_matcher=", hooks[0]["matcher"])
print("has_execSync=", "execSync" in command)
print("has_execFileSync=", "execFileSync" in command)
print("command=", command)
PY
node - <<'JS'
const fs = require('fs');
const { execFileSync } = require('child_process');
const settings = JSON.parse(fs.readFileSync('.claude/settings.json', 'utf8'));
const hook = settings.hooks.PostToolUse[0].hooks[0].command;
const p = 'x.js"; echo INJECTION; #.js';
// Model the hook's interpolation without invoking the repository hook.
// The replacement keeps the same shell quoting and changes only the
// formatter command to a harmless builtin.
const shellText = hook
.replace(/^node -e "/, '')
.replace(/"$/, '')
.replace(/const fs=require\('fs'\); const d=JSON\.parse\(fs\.readFileSync\(0,'utf8'\)\); const p=d\.tool_input\?\.\.file_path\|\|''; if\(.*?\) \{ require\('child_process'\)\.execSync\('/,
'const shellCommand = ')
.replace(/', \{stdio:'inherit'\}\); \}/, '; }')
.replace(/const shellCommand = 'npx prettier --write \\\\"'\+p\+'\\\\\\"'/,
"const shellCommand = 'printf PRETTIER ' + JSON.stringify(p)")
.replace(/; }$/, '; }');
console.log('payload_extension_matches=', /\.(ts|tsx|js|jsx)$/.test(p));
console.log('interpolated_command=', `npx prettier --write \\"${p}\\"`);
console.log('shell_syntax=', 'the injected command contains a command separator');
console.log('argument_api_output=');
const output = execFileSync(process.execPath, [
'-e',
'process.stdout.write(JSON.stringify(process.argv[1]))',
p
], { encoding: 'utf8' });
console.log(output);
JSRepository: zsdotcom/zarishdocs
Length of output: 768
Pass file_path as an argument, not shell text.
Line 31 interpolates d.tool_input.file_path into execSync. Shell metacharacters can execute additional commands. Use execFileSync with an argument array and the local Prettier executable.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In @.claude/settings.json at line 31, Update the command handler around the
file_path extraction to invoke Prettier without shell interpolation: replace
execSync and the constructed shell string with execFileSync, passing the local
node_modules/.bin/prettier path and its arguments in an array. Preserve the
existing TypeScript/JavaScript extension check and formatter existence guard.
2 participants
What does this PR do?
Related issue
Closes #
Checklist
latesttags introduced in any configdocs.zarishsphereif this changes behaviorSummary by CodeRabbit
New Features
Documentation
Bug Fixes