Back to catalog / Agent workflows
Ground an architecture map in an unfamiliar repository
Compare source code with stale architecture docs in a disposable repo, then add evidence to the live session.
Goal
Produce a file-grounded architecture note that flags documentation contradicted by implementation.
Steps
- Inspect the tree, then use
glob,grep, andreadto trace actual entry points and service boundaries. [explore] - Compare each documented claim with source files; mark unverified claims unknown instead of repeating stale
ARCHITECTURE.md. [explore] - Write findings with paths to
/tmp/NOTES.md; add newly found evidence to the running session withsessions.resources.add(session_id=..., type="file", ...). [explore]
How to verify
Confirm service layout against source directories, entry files, and call paths; verify added resources appear in the current session.
Troubleshooting
If the answer repeats stale docs, trace callers in code and record the contradiction; attach missing evidence before continuing.
Permissions
Prerequisites
- Python 3.11+, an Anthropic API key, the Managed Agents beta SDK, and a disposable repository fixture.
Permissions
- Mount only approved code; exclude secrets, production data, and personal files.
Limitations
The sample `/tmp/NOTES.md` is sandbox-local; a model summary does not replace maintainer review.
Pinned upstream source
- Repository
- anthropics/anthropic-cookbook
- File path
- managed_agents/CMA_explore_unfamiliar_codebase.ipynb
- Pinned commit
9c1e493ddd2b82df399d2a61b9c806c539e18be6- Source checked
- 2026-10-14 · Reviewed against the pinned upstream source
Repository metrics
These values describe the upstream repository, not popularity or likes for an individual skill or playbook.
- Stars
- Not collected
- Forks
- Not collected
- Observed at
- Not collected
No metrics snapshot is available yet.
License: MIT
Main topic
Take a coding Agent change from context to verification