gsc-scan
Skill17 starsChanged yesterdayApache-2.0View on GitHub
Use when scanning a repo or PR with GSC and reading the coverage-aware report.
In this repository4 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- gsc-fix-verifySkill
- gsc-triageSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 54
Read scan_status first: ok or partial.
- Distinctive: line 55
Read the coverage block before trusting any verdict:
- Very specific: line 68
Do not report "clean" on an empty finding list while detectors_not_run is non-empty — that is an incomplete scan, not a pass.
- Very specific: line 74
No PoC attached ≠ false positive: several classes (e.g.
≈ “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 Scan
When to use
You are asked to scan a repository, a local path, or a pull request diff for
vulnerabilities, secrets, dependency issues, or IaC misconfigurations with GSC,
and to interpret the resulting report. GSC is a SAST engine with PoC
verification; the engine runs from the Docker image
ghcr.io/poliakarmai/gsc-scanner:v1 (BUSL 1.1), the Action poliakarmai/gsc@v1
is Apache 2.0.
Commands
GitHub Action — one file, zero config (.github/workflows/gsc.yml):
name: GSC Audit
on:
pull_request:
branches: [main, master] # must match your default branch, else it never fires
jobs:
audit:
runs-on: ubuntu-latest
permissions:
contents: read
pull-requests: write
steps:
- uses: actions/checkout@v4
- uses: poliakarmai/gsc@v1
with:
fail_on_critical: false
# fail_on_score: 80 # fail the PR when the score drops below this
# sarif: true # also emit SARIF for GitHub Code ScanningCLI in Docker:
docker run --rm -v "$PWD:/repo:ro" -w /repo \
ghcr.io/poliakarmai/gsc-scanner:v1 scan . --ci --jsonUseful Action inputs: path (default .), deep_scan (default true),
fail_on_critical, fail_on_score (0 = off, 1–100), max_findings_to_comment
(default 10), sarif, reachability, and BYO-LLM llm_api_key / llm_base_url /
llm_model (without a key the engine runs regex-only).
How to read the result
- Read
scan_statusfirst:okorpartial. - Read the coverage block before trusting any verdict:
files_scanned / files_total / files_truncated— how much of the tree was seen;detectors_total / detectors_ran / detectors_not_run— which detectors executed.
scan_status: partial(a cap hit, not a crash) means a low finding count is not a clean repo. Say which detectors did not run.- Per finding, check whether a PoC is attached. Two layers: a deterministic
template (works without an LLM key, no confidence gate) and an LLM PoC for
findings with
confidence >= 0.80. verify_findinganswersstatus: "unavailable"when the sandbox cannot start or no PoC is attached. That is not a verdict about the finding.
Pitfalls
- Do not report "clean" on an empty finding list while
detectors_not_runis non-empty — that is an incomplete scan, not a pass. pr-gate/ diff mode only sees changed files; it can never give a whole-repo verdict. Use a full scan for that.scan_repois bounded byGSC_SCAN_TIMEOUT(default 300 s) andGSC_MAX_FILES(default 50 000). A truncated scan is not persisted.- No PoC attached ≠ false positive: several classes (e.g. hardcoded secrets) have
no deterministic template, and
UNAVAILABLEis the honest answer.
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.

