Repo Cartographer
One-command Markdown map of a repository (languages, layout, manifests, entry points, symbols) sized to a character budget.
Not recommended on the current evidence, whatever your baseline: success with the capability is not shown to be at least as good (difference interval [-0.3321, 0.1641], margin 0.05).
- Id
ak:agentickeychain:repo-cartographer:0.1.0- Publisher
- Agentic Keychain (first-party: run by the registry's operator, Vayu Solutions; first-party skills get no ranking advantage)
- State
- tested
- Price
- USD 5.00 one-time (one-time price in USD; not on sale, see Availability)
- License
- LicenseRef-AK-Paid-Skill
- Availability
- checkout for this capability is not configured yet
- Decision
do_not_recommend(the API's answer when no task is given; your baseline does not change it)- Trust score
- 0.95 (provenance checks, not quality; see Method)
- Models measured
- claude-sonnet-5
- Machine-readable
- JSON, Markdown
What it does
Gives an agent a compact map of an unfamiliar repository before it starts opening files, so it spends fewer tool calls and input tokens on orientation. The map lists languages by lines of code, the directory layout, build manifests, entry points and the main symbols per file, and stays inside a character budget the agent chooses.
When to use it
- The repository has at least 50 files and you have not mapped it yet.
- You need to find where a feature lives in a codebase you have not read before.
When not to use it
- The repository is small enough to list and read directly.
- The task concerns one file you already know.
- You cannot run Python 3.8 or newer in the environment.
Free alternatives
- git ls-files plus the README — Lists every tracked file; enough for small repositories.
- universal-ctags — Full symbol index for many languages; needs installing and gives no layout or entry-point summary.
Evidence
| Label | Model | Pairs | Saving per run (95% CI) | Cost reduction (median) | Success without → with | Fixtures |
|---|---|---|---|---|---|---|
KEYCHAIN TESTED | claude-sonnet-5 | 30 | USD -0.0269 [USD -0.0531, USD -0.0008] | -1% | 97% → 87% | synthetic |
Measured on the registry's fixtures. On this evidence the API answers do_not_recommend whatever your baseline: success with the capability is not shown to be at least as good (difference interval [-0.3321, 0.1641], margin 0.05).
Limitations the API states with these figures:
- The figures were measured on claude-sonnet-5; for another model they are for reference only.
- Benchmarked on synthetic fixtures with planted answers; treat the savings as an upper bound.
- Install and review time is not estimated.
Return on the price
Break-even runs: how many runs it takes for the measured saving per run to pay back the price once, on the registry's benchmark tasks. The saving is taken at the lower bound of its 95% interval. Your own return depends on what one run costs you now; resolve_task and evaluate_capability compute it for your task.
| Model | Label | Pairs | Success at least as good | Break-even runs at USD 5.00 one-time |
|---|---|---|---|---|
| claude-sonnet-5 | KEYCHAIN TESTED | 30 | not shown | not shown: success is not shown at least as good |
All evidence and attestations · Provenance · Changelog
Permissions it needs
- Network
- no
- Shell
- yes
- Credentials
- no
- Files
- read
Get it
Not offered: checkout for this capability is not configured yet.
Package
- File
repo-cartographer-0.1.0.zip- SHA-256
649c83e7a9fca36fecd6c04921139a3fd732412782716c89bf0ec5df9d85f8dc- Size
- 7,606 bytes
- Files
- 2