avoid-empty-review-artifact-evidence-in-plan-redrafts
Keep plan redrafts truthful during adversarial rerun loops by never citing zero-byte or not-yet-generated review artifacts as resolved evidence.
Best use case
avoid-empty-review-artifact-evidence-in-plan-redrafts is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Keep plan redrafts truthful during adversarial rerun loops by never citing zero-byte or not-yet-generated review artifacts as resolved evidence.
Teams using avoid-empty-review-artifact-evidence-in-plan-redrafts should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/avoid-empty-review-artifact-evidence-in-plan-redrafts/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How avoid-empty-review-artifact-evidence-in-plan-redrafts Compares
| Feature / Agent | avoid-empty-review-artifact-evidence-in-plan-redrafts | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Keep plan redrafts truthful during adversarial rerun loops by never citing zero-byte or not-yet-generated review artifacts as resolved evidence.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Avoid Empty Review Artifact Evidence in Plan Redrafts
Use when redrafting a plan after Codex/Gemini/Codex MAJOR findings and preparing a fresh adversarial rerun.
## Problem
A common failure mode in plan-redraft loops is updating the plan text to say the latest MAJOR findings were "addressed" and citing same-day review artifact paths before the rerun has actually completed. If those artifact files are empty or placeholders at review time, the next reviewer correctly flags the plan as evidence-weak or internally contradictory.
## Rule
Do not claim a fresh review wave is resolved unless the corresponding review artifacts:
1. exist on disk
2. are non-empty
3. actually contain the cited verdict/findings
Zero-byte files, placeholder files, or paths that will only be populated later do not count as evidence.
## Recommended wording
If you have redrafted the plan but have not yet completed the rerun, use wording like:
- "fresh adversarial rerun pending"
- "latest redraft prepared for rerun"
- "prior MAJOR findings addressed in draft; verification pending fresh review"
Avoid wording like:
- "Codex MAJOR addressed"
- "Gemini findings resolved"
- "latest review confirms ..."
unless the new artifact is already present and non-empty.
## Practical workflow
1. Redraft the plan against the last completed review artifacts.
2. In `## Adversarial Review Summary`, keep the historical verdicts truthful.
3. If describing the new draft, explicitly mark rerun status as pending.
4. Run the fresh provider review.
5. Verify the new review files are non-empty.
6. Only then update the summary to say the new review wave produced MINOR/APPROVE/MAJOR and whether the plan is approval-ready.
## Verification pattern
Before citing a review artifact in the plan body or summary, check:
- `wc -c scripts/review/results/<file>.md`
- `read_file()` or `terminal("sed -n '1,40p' ...")`
- `git status -s scripts/review/results/` to catch deleted/renamed tracked stubs or untracked real artifacts
- `git ls-tree -r --name-only HEAD scripts/review/results | grep <issue>` after commit to verify the committed evidence paths match the plan's header / Artifact Map
Minimum acceptable evidence:
- a verdict line
- a summary line
- at least one substantive finding or explicit statement of what was checked
- the exact cited artifact path is non-empty and committed when requesting user approval
## Round-archiving rule for mutable fanout outputs
If the review fanout writes mutable provider paths like `YYYY-MM-DD-plan-<issue>-AGENTS.md`, do not `mv` a valid artifact away and then cite the now-empty/missing mutable path in the plan. Prefer one of these safe patterns:
1. Copy the populated mutable file to an immutable round path (`...-Codex-r2.md`) after verifying it is non-empty, then update the plan to cite the immutable path; or
2. Keep citing the mutable path, but do not move it and verify it remains populated after any rerun.
If a rerun fails and leaves `*.md` as 0 bytes with only a `*.err` sidecar, treat that run as provider-infrastructure evidence only. Either rerun successfully or cite an explicit unavailable/waiver artifact; never let the plan's front-matter or Artifact Map point at the 0-byte file.
## Why this matters
This prevents self-inflicted review failures where the plan appears to claim evidence that does not yet exist. It keeps approval-stage artifacts honest and avoids wasting a rerun on governance contradictions instead of real plan quality.Related Skills
plan-review-prompt-refresh-after-plan-edits
Refresh reviewer prompt files from the latest on-disk plan before every adversarial re-review. Prevents Codex/Gemini from critiquing stale plan text after local edits.
workspace-hub-overnight-plan-monitor
Monitor and reconcile workspace-hub overnight planning or implementation batches, including process status, result artifacts, issue/commit verification, and controlled failed-lane recovery.
plan-gated-issue-execution-wave
Execute a multi-issue architecture/planning wave in a plan-gated repo, then safely transition approved issues into implementation with file-based Codex prompts, local approval markers, subprocess monitoring, and cleanup handling for sandbox/hook edge cases.
mixed-ops-vs-repo-fix-plan-boundary
Plan mixed operational-vs-repo remediation issues by proving live-state classification first, then only proposing code changes for confirmed repo-owned failure paths.
wave-based-parallel-plan-execution
Orchestrate phase execution by discovering dependencies, grouping into waves, spawning subagents, and collecting results with optional wave filtering
tdd-verification-and-adversarial-review
Verify pre-written TDD tests pass, conduct adversarial code review on committed diffs, and route findings to existing issues
plan-gated-issue-validation-workflow
Systematic validation pattern for plan-approved GitHub issues with pre-existing deliverables
plan-gated-issue-implementation
Workflow for executing pre-approved GitHub issues with mandatory validation checkpoints
multi-role-agent-contract-review-pipeline
Execute a 4-role agent team (Planner/Architect/Reviewer/Integrator) pipeline for self-reviewing knowledge artifacts before delivery
gsd-adversarial-review-pattern
Catch hidden test failures by running adversarial review on sparse-data edge cases before final push
adversarial-code-review-tdd
Systematic adversarial review pattern to identify breaking assumptions in already-passing test suites
adversarial-code-review-for-committed-diffs
Systematic process for reviewing already-committed code changes to catch type inconsistencies, edge cases, and docstring gaps