# Export Plan Review Comments > Export the survived comments from a completed implementation-plan review into a timestamped Markdown findings artifact. Use when the user asks to save, write, or export plan-review comments/findings under `.ai/impl_plans/PLAN_ID/findings/REVIEW_TIMESTAMP/finding.md` while preserving the original review wording and wrapping each finding in paired `comment` markers. - Skill: `blockscout/export-plan-review-comments` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add blockscout/export-plan-review-comments` - Raw SKILL.md: https://api.skillmd.com/api/skills/blockscout/export-plan-review-comments/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: blockscout (https://skillmd.com/u/blockscout) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/blockscout/export-plan-review-comments --- # Export Plan Review Comments Export the final review comments without re-reviewing the plan, changing verdicts, or implementing anything. ## Resolve the source and target 1. Identify exactly one plan file from the user request or the completed review: ```text .ai/impl_plans/.md ``` If more than one plan is plausible, ask instead of guessing. 2. Resolve the review timestamp from the exact scratchpad paths cited by the comments: ```text .ai/impl_plans//scratchpads//... ``` Reuse that timestamp verbatim. Do not generate the current time, glob scratchpad directories, or reuse a timestamp from another review. If no exact review timestamp is recoverable, ask the user. 3. Write to: ```text .ai/impl_plans//findings//finding.md ``` 4. Treat every item actually present in the final review's actionable-comments section as one survived finding. Do not export summaries, coverage notes, closed candidates, intermediate commentary, or test-checklist repetitions. ## Preserve each comment - Keep the original review wording almost verbatim. - Do not normalize findings into a new schema. - Do not rewrite links, severity labels, verdicts, recommendations, rationales, action text, or scratchpad paths. - Make only a tiny edit when required to make a finding understandable on its own. - Keep the original numbering text inside the block body. - Preserve Markdown lists and blank lines within a comment. Wrap each finding in a separate region: ```markdown 1. **Severity:** Major **Location:** ... **Problem:** ... **Recommendation:** ... **Rationale:** ... **Scratchpad:** ... ``` Use a unique lowercase kebab-case slug. The begin and end slugs must match. Keep only blank lines between regions and no document-level introduction or summary. ## Write safely - Create only the timestamped findings directory and `finding.md`. - Use `apply_patch` for the file content. - Do not modify the implementation plan, scratchpads, application code, tests, or documentation. - If the target already exists with different content, do not overwrite it unless the user explicitly requested replacement or update. ## Validate Run the bundled validator with the exact number of exported review comments: ```bash python3 .agents/skills/export-plan-review-comments/scripts/validate_comments_export.py \ .ai/impl_plans//findings//finding.md \ --expected-count ``` Fix every validation error before finishing. Then read the file once to confirm that wording and links still match the source comments. ## Respond Reply with only a clickable link to the created file: ```markdown [.ai/impl_plans//findings//finding.md](.ai/impl_plans//findings//finding.md) ```