gsc-fix-verify
Skill17 starsChanged yesterdayApache-2.0View on GitHub
Use when fixing a GSC finding and proving it with verify_fix and a re-scan.
In this repository4 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- gsc-scanSkill
- gsc-triageSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 41
After the verdict, re-check the report: the finding is gone, no new finding of the same class appeared elsewhere (common when a fix is …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
GSC Fix & Verify
When to use
You are asked to fix a specific GSC finding or the blocking set, and then to prove the fix — not merely to make the finding disappear from the report.
Commands
# 1. Inspect
get_finding # location, class, evidence, finding_key
get_poc # the reproduction: deterministic template or LLM PoC
# 2. Edit the code so the reproduction no longer works.
# 3. Prove
verify_fix # re-scan after the patch: fixed | still_vulnerable | inconclusive
verify_finding # fallback when verify_fix has nothing to compare for this class
# 4. Confirm no regression (whole repo, not just the patch)
docker run --rm -v "$PWD:/repo:ro" -w /repo \
ghcr.io/poliakarmai/gsc-scanner:v1 scan . --ci --jsonIn CI the operational definition of done is: the Action step exits zero with
your configured fail_on_critical / fail_on_score.
How to read the result
verify_fix: fixed= the detector no longer fires on a re-scan (GSC-003). That is a re-scan verdict, not an executed-exploit verdict — say it that way.verify_fix: still_vulnerable= the finding survives; do not claim the fix landed.verify_fix: inconclusive/unavailable= no evidence either way. Checkdoctor(sandbox, state volume, allowed roots) and re-run; do not convert it into "fixed" by wishful reading.- After the verdict, re-check the report: the finding is gone, no new finding of
the same class appeared elsewhere (common when a fix is copy-pasted), and
detectors_not_runis empty or explicitly noted as a residual blind spot.
Pitfalls
- Fixing the pattern without fixing the behaviour (base64-ing a secret instead of loading it from config) passes the detector and leaves the bug. Verify semantics, not just the disappearing finding.
- For credentials, "fixed" means removed and rotated — removing the string from a file does not un-leak a key that was committed.
- A green diff-scoped gate (
scan_diff) is not a green repo: re-scan fully after a fix. --read-onlyon the state volume is unsupported; findings would come back withpersisted: falseandverify_fixwould have nothing to compare.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

