agentleFS
midFangmidFang/ai-agent-skills-workflow/plan-csv-execute/SKILL.md

plan-csv-execute

Skill2 starsChanged 5 months agoMITView on GitHub

Execute a single issues CSV snapshot as the unique task boundary and state source, then close the loop through implementation, review, validation, and local git commit without push. Use when the user explicitly mentions `$plan-csv-execute`, says "按 issues csv 执行", "根据 CSV 闭环开发", or wants to replace the deprecated `/prompts:issues_csv_execute` workflow.

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 6 are distinctive, and 9 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this skill as the replacement for the deprecated custom prompt /prompts:issues_csv_execute.

    line 8
  2. Distinctive:

    Treat the user-supplied CSV as the only state source.

    line 21
  3. Distinctive:

    Do not create or sync any extra summary CSV files unless the user explicitly asks.

    line 22
  4. Distinctive:

    Default goal: move every unfinished row to a closed local state.

    line 23
  5. Very specific:

    Use update_plan for multi-step execution runs.

    line 31
  6. Distinctive:

    If the CSV does not match, stop and ask the user to regenerate or repair the CSV first.

    line 41
  7. Very specific:

    Write a short picked_reason: note into the row when selecting it.

    line 47
  8. Very specific:

    If acceptance_criteria, review requirements, test_mcp, or refs are incomplete, update the CSV first.

    line 52
  9. Very specific:

    Read refs, gather the minimum code context, and implement only the work inside that row's boundary.

    line 54
  10. Distinctive:

    Run the most relevant tests or checks you can.

    line 55
  11. Very specific:

    If full validation is not possible, record all of these in notes:

    line 56
  12. Very specific:

    Commit code-only changes with a message exactly equal to the row's git_message.

    line 65
  13. Very specific:

    If the commit succeeds, set git_state=已提交.

    line 66
  14. Very specific:

    If the commit fails, set git_state back to 未提交 and record blocked:git commit failed <reason> in notes.

    line 67
  15. Distinctive:

    Stop when all remaining rows are blocked and summarize the blockers plus the smallest next decision the user must make.

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Issues CSV Execute

Use this skill as the replacement for the deprecated custom prompt /prompts:issues_csv_execute.

Invocation

Preferred explicit trigger:

  • $plan-csv-execute issues/2026-03-31_10-10-00-login-flow.csv

Natural-language triggers:

  • 按这份 issues csv 执行
  • 根据 CSV 闭环开发直到提交

Core Rules

  1. Treat the user-supplied CSV as the only state source.
  2. Do not create or sync any extra summary CSV files unless the user explicitly asks.
  3. Default goal: move every unfinished row to a closed local state.
  4. Each issue must be handled as a full loop:
    • implement
    • review
    • validate
    • local git commit
  5. Never push.
  6. By default, do not add issues/*.csv, plan/, *.md, or other process documents to commits.
  7. Use update_plan for multi-step execution runs.

Required Header

Expect this exact header:

id,priority,phase,area,title,git_message,description,acceptance_criteria,test_mcp,review_initial_requirements,review_regression_requirements,dev_state,review_initial_state,review_regression_state,git_state,owner,refs,notes

If the CSV does not match, stop and ask the user to regenerate or repair the CSV first.

Issue Selection Order

  1. Finish hanging work first: rows with git_state=未提交 and dev_state=进行中|已完成.
  2. Then prefer unfinished rows by P0 -> P1 -> P2.
  3. Write a short picked_reason: note into the row when selecting it.

Per-Issue Workflow

  1. Re-state the target row: id, title, acceptance criteria, and main risks.
  2. If acceptance_criteria, review requirements, test_mcp, or refs are incomplete, update the CSV first.
  3. Set dev_state=进行中 and review_initial_state=进行中.
  4. Read refs, gather the minimum code context, and implement only the work inside that row's boundary.
  5. Run the most relevant tests or checks you can.
  6. If full validation is not possible, record all of these in notes:
    • validation_limited:<reason>
    • manual_test:<steps>
    • evidence:<what was verified>
    • risk:<low|medium|high> <why>
  7. Once the issue is genuinely ready, set:
    • dev_state=已完成
    • review_initial_state=已完成
    • review_regression_state=已完成 when the available regression evidence is sufficient
  8. Commit code-only changes with a message exactly equal to the row's git_message.
  9. If the commit succeeds, set git_state=已提交.
  10. If the commit fails, set git_state back to 未提交 and record blocked:git commit failed <reason> in notes.

Stop Conditions

  • Continue until every row is closed, or
  • Stop when all remaining rows are blocked and summarize the blockers plus the smallest next decision the user must make.

Reply Shape

For each completed loop, report:

  • id/title
  • key changed files with path:line
  • tests or checks actually run
  • validation limits, if any
  • local commit hash
  • remaining blockers or next issue count

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.