Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →fedoroff-vlad/map-project
Use at the START of onboarding or documenting an unfamiliar project — to scan and index it before writing a line of docs. Fires on: "просканируй проект", "проиндексируй репозиторий", "разберись в проекте X", "map the project", "scan the repo", "что это за проект и из чего он состоит". Runs a mechanical scanner, then reads the flagged files to produce a durable project map: modules, build system, toolchain, deploy manifests, services/ports, env keys, entry points — written to docs/onboarding/proj
Passed 11 of 18 test cases with the skill, and 4 of 18 without it.
and 9 more in the full report
10 fixed4 passed either way7 still failing1 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
We are running on Linux with the agent skills available at `tools/agent-skills`. Write the command line to scan the target repository located at `/var/repos/my-app`.
| Checked | Without skill | With skill |
|---|---|---|
| Executes scan-project.sh using shell execution with the target root path parameter. | Missed | Passed |
The instructions your model reads when the skill loads.
The first phase of onboarding. Produce ONE artifact — docs/onboarding/project-map.md
— that every later phase (document-project, run-guide) reads instead of
re-deriving the repo from scratch. It is written to disk, not held in
context: a fresh session after a context reset re-reads it and continues.
The map is agent-facing — write it in English always (like the
journey-log and spec skeletons), regardless of ONBOARDING_LANG. Only the
human-facing READMEs/RUN.md follow the reader's language.
The scanner emits candidates with file:line — it never decides meaning.
You (the agent) open the flagged files and extract the semantics. Never copy a
scanner hit into the map as fact without reading the source line.
If project-map.md already exists, this is a refresh: re-run the scanner,
diff against the current map, update only the rows that changed, and note removed
/ added modules. Do not discard hand-edited notes in the map — merge over them.
The scanner's "Sensitive files" section lists paths only and never opens
them. If it flags anything (a real .env, *.pem, id_rsa, keystore, …):
do not read the file, record only its path + that it exists, and warn the
user a secret may be committed. For a public repo, hand off to
scrub-identity before anything is written or pushed. A secret value must never
reach the map, the journey log, or your output.
Locate the target. Get the project root (by name or path). Everything below runs against it, not against this skills repo.
Run the scanner (pick the platform twin; both are multi-platform):
# macOS / Linux / Git-Bash
test cases passed, vs 4 / 18 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.