Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →boshu2/codebase-recon
Reconstruct a repository as cited
Passed 12 of 17 test cases with the skill, and 4 of 17 without it.
and 12 more in the full report
10 fixed5 passed either way7 still failing0 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 the validation utility script against a codebase recon manifest located in .agents/scratch/codebase-recon/run-123/codebase-recon.json, but the target repository we analyzed is located in an external directory /var/workspace/target-repo. Which command line flag must be passed to skills/codebase-recon/scripts/validate-output.sh so evidence checks resolve against the analyzed repository?
| Checked | Without skill | With skill |
|---|---|---|
| Command invocation passes the --repo-root option to specify target directory | Missed | Passed |
The instructions your model reads when the skill loads.
Build a reusable, falsifiable model of a repository. This skill reports what the tree and executable probes support; it does not edit code or issue a final PASS/WARN/FAIL verdict.
codebase-recon --mode=baseline --view=audit --lens=cli --depth=standard
on agentops-wt/train2-c: trace the `ao gate check` entry point from
cli/cmd/ao/gate_composition.go through to its tests. I'm onboarding and need a cited
mental model, not a file inventory.
Observable in the trace, without reading the prose:
.agents/scratch/codebase-recon/<run-id>/codebase-recon.json names the
exact commit and passes
skills/codebase-recon/scripts/validate-output.sh <codebase-recon.json>.fact cites file:line, and every inference cites the facts it
rests on.lens (persistence, auth, CLI, build, or test) follows one
entry-to-test flow to completion, or the report names the exact
file:line where the trace was cut.codebase-recon.md.test cases passed, vs 4 / 17 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.