Works with the AI Agents you already use
repoctx
Keep AI agent context files in sync with your code by detecting stale paths and regenerating command tables.
Free
repoctx
Example session with this skill installed
The project structure changed after the refactor. Audit the agent context files to see if anything is broken and refresh the command tables.
- Read your context and instructions
- Compiled the repoctx
I have audited the repository context files. I found one stale path in AGENTS.md and have regenerated the command tables to reflect the new scripts in your manifest.
Context health
File: AGENTS.md · Score: 85/100
| Check | Result |
|---|---|
| commands | 8 claimed, all found |
| paths | 12 referenced, 1 stale |
What to fix
src/legacy/utils.ts— referenced in the context file, but the file was deleted or moved during refactor.
How to fix
- Update the path in the human-readable section of
AGENTS.md, or runrepoctx generate .if the command list changed.
# Regenerating code-derived tables...
repoctx generate .
# Final verification...
repoctx audit . --check
Next steps
- Review the human-written "Conventions" section in
AGENTS.mdto ensure no manual instructions reference the oldsrc/legacypath. - Commit the updated
AGENTS.mdto ensure sub-agents stay synced with the new structure.
Connects securely to your tools. The creator never sees your data.
About this skill
Context rot kills agent performance. When your AGENTS.md or .cursorrules files claim commands that no longer exist or point to moved paths, your AI model hallucinates and fails. This skill keeps your agent instructions truthful by syncing them with your actual codebase.
What it does
- Detects context rot by identifying ghost commands and stale file paths in your agent instructions.
- Regenerates code tables to automatically update command lists and module dependencies from your manifests.
- Preserves human content by only modifying sections between specific comment markers.
- Enforces context health via CI/CD gates that fail when documentation drifts from the implementation.
- Bootstraps new repos with a standardized context file if one does not exist.
How it works
- Analyze: Run
repoctx infoto see what manifests, commands, and modules are currently detected in the repository. - Sync: Execute
repoctx generateto refresh all code-derived tables while keeping your manual notes intact. - Verify: Use
repoctx audit --checkto validate that all referenced files and commands actually exist on disk. - Automate: Integrate
audit --sinceinto your PR pipeline to catch drift introduced by new changes without failing on legacy debt.
Frameworks & tools
Requires the repoctx CLI installed via npm or Homebrew. It works with any repository containing standard manifests and supports AGENTS.md, CLAUDE.md, and .cursorrules files.
Why this beats prompting it yourself
Manual updates to agent context files are often forgotten during refactors, leading to "ghost" instructions that mislead the AI. This skill provides a systematic, CLI-backed way to verify facts, ensuring the agent always operates on a valid map of the repository.
Use cases
- Syncing
AGENTS.mdafter refactoring scripts inpackage.jsonorpyproject.toml. - Cleaning up stale file references in
.cursorrulesafter moving documentation folders. - Gating pull requests to ensure developers update agent instructions alongside code changes.
- Auditing inherited codebases to see if the existing agent documentation is still trustworthy.
Known limitations
Requires specific repoctx comment markers to modify files. It cannot update human-written sections like project overviews or conventions and requires a manifest file to extract data.
Requires repoctx CLI (npx/brew). Must have a manifest file (package.json, etc.). Only updates content within repoctx comment markers. Cannot auto-fix human-written descriptions.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 25 days ago
- Free with an Agensi account
Requires