Skill

integrating-sast-into-github-actions-pipeline

skill:mukul975/anthropic-cybersecurity-skills#integrating-sast-into-github-actions-pipeline

Integrates CodeQL and Semgrep SAST scanning into GitHub Actions, covering scans on pull requests/pushes, rule tuning to cut false positives, SARIF upload to GitHub Advanced Security, and merge-blocking quality gates for high-severity findings. Use when adding automated code vulnerability detection to CI, enforcing consistent SAST org-wide, or producing SOC 2/PCI DSS/NIST SSDF compliance evidence.

SkillsMP

Verdict

Suspicious

Risky behavior the analysis could not rule out.

Analyzed Oct 27, 2026, 21:46 UTC at 54a7988

findings

medium

highest severity

threat classes

4,532

files scanned

190/190

checks satisfied

package advisories

Why this verdict

suspicious: finding skill:skills/integrating-sast-into-github-actions-pipeline:skills/integrating-sast-into-github-actions-pipeline/SKILL.md:119:SUPPLY

Security findings 1

Supply chain

mediumContextual riskmedium confidence

The skill's CI workflow templates instruct repositories to execute the mutable container image semgrep/semgrep:latest and the tag-pinned (not SHA-pinned) third-party action returntocorp/semgrep-action@v1 on every scan, so a hijacked tag or image would run attacker-controlled code in CI with access to the checked-out source and the SEMGREP_APP_TOKEN secret.

skills/integrating-sast-into-github-actions-pipeline/SKILL.md:119-121

image: semgrep/semgrep:latest

vector:instruction

Counter-argument considered

Every factual element of the claim is verbatim-supported: the skill body and the asset template both set the mutable image semgrep/semgrep:latest, the reference file and the agent's gen-workflow output use tag-pinned returntocorp/semgrep-action@v1, and SEMGREP_APP_TOKEN is injected into the step executing inside that container. No SHA pinning or mitigation exists anywhere in the shown units, the unpinned nature defeats the pinned-official-registry-install carve-out, and no declared claim (only D2 on custom rule authoring) covers or qualifies this behavior, so the contextual supply-chain risk stands.

Open questions 2

What the analysis could not settle from the code alone.

Both detector leads (DET:ebeff87abde6923f, DET:94ae82e57d0d2a56) cite the raw.githubusercontent.com URL in the SARIF aggregation script; verification against SRC:dba1d89a2cc8f6ba and SRC:2f5f4ae305b306fd shows it is only a literal $schema identifier string written by jq into local output, with no fetch or send instruction, so both leads are unconfirmed and no EXFIL finding is raised. No open question remains.

Frontmatter tags the skill with offensive supply-chain ATT&CK techniques T1195 and T1554 plus T1059.004 (Unix Shell), which do not obviously match the defensive SAST purpose; determine whether this is domain/coverage tagging or an intent signal.

Supply chain