- Home
- Skills
- Technical Documentation
- Readme Keeper
Works with the AI Agents you already use
Readme Keeper
Verify and repair stale README claims by cross-referencing commands, paths, and versions against the repository.
Free
Readme Keeper
Example session with this skill installed
Run an audit on the current README.md. The setup steps seem broken for new hires. Verify all commands and paths against the repo and give me a report of what to fix.
- Read your context and instructions
- Compiled the readme keeper
I have audited README.md against the repository filesystem and manifests. I found 3 failing claims where the documentation has drifted from the actual code.
README Audit Report
| Claim | Type | Evidence | Status |
|---|---|---|---|
scripts/setup.sh | Path | File not found at root | FAIL |
npm run dev:legacy | Command | Not found in package.json scripts | FAIL |
Node >= 18.0.0 | Version | .nvmrc specifies v20.11.0 | FAIL |
docs/architecture.md | Link | File exists | PASS |
python3 | Command | Found on PATH | PASS |
LICENSE | Path | File exists | PASS |
Findings and Corrections
Broken Path: The README references scripts/setup.sh, but the file was moved to .build/setup.sh. Update the Quickstart section to reflect the new location.
Stale Command: The command npm run dev:legacy no longer exists. The current development command is npm run dev.
Version Mismatch: The README claims Node 18 is required, but the project configuration now enforces Node 20. Update the requirements section.
Next steps
- Update
README.mdlines 12, 45, and 82 with the corrected paths and commands. - Run
python3 scripts/readmecheck.py check README.mdto verify the fixes. - Commit the updated README alongside the recent script changes.
Connects securely to your tools. The creator never sees your data.
About this skill
READMEs rot faster than code. A script gets renamed, a dependency version bumps, or a quickstart step is forgotten, leaving new developers stuck in configuration hell. This skill stops README drift by treating documentation as testable code, ensuring every claim made to a user is backed by the actual state of the repository.
What it does
- Claim extraction identifies every command, file path, version requirement, and link within your documentation.
- Automated verification cross-references extracted claims against the local filesystem, manifest files, and environment paths.
- Drift reporting generates a detailed table of evidence showing exactly where a README lies compared to the repo.
- Structure enforcement provides an optimized "newcomer-first" outline based on the specific features detected in your project.
- CI integration enables an exit-code-based check to prevent merging PRs with broken documentation links or invalid setup commands.
How it works
- Run
claimsto inventory every testable statement, from bash snippets to version strings. - Execute
checkto validate those claims againstpackage.json,go.mod, file structures, and PATH. - Use the generated
reportto identify specific sentences that require correction based on the evidence. - Apply the
outlinecommand when starting new docs to ensure mandatory sections like "What this does not do" are included.
Frameworks & tools
The skill utilizes a standalone Python utility (readmecheck.py) compatible with any repository structure including Node.js, Python, Go, and Rust. It integrates with standard CI/CD pipelines via shell commands.
Why this beats prompting it yourself
Generic LLMs hallucinate setup steps based on common patterns rather than your specific files. This skill tethers the agent to a local verification script that proves a command exists before suggesting it, eliminating the "it works on my machine" documentation trap.
Use cases
- Auditing a legacy repository where the setup steps no longer work on a fresh clone.
- Verifying all internal anchors and file references before a major version release.
- Generating a technically accurate README for a new project that reflects the actual file structure.
- Preventing documentation regression by adding a verification step to a GitHub Action.
Known limitations
Does not execute commands to verify their output; it only verifies the command is invokable. External URLs are not pinged. Does not verify claims that lack a corresponding file-based manifest.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
1 install
Downloaded by developers to date
Free forever
No account required to browse
Trust & safety
Security scanned
Verified clean 4 days ago
- Free to download with an account
Requires