semgrep-analyze

Semgrep Analyze

Role: worker. This skill runs Semgrep and reports findings — it does not fix code.

You have been invoked with the /semgrep-analyze skill. Run a Semgrep scan and return structured findings.

Constraints

1. Do not modify code. Report findings only.
2. Return structured JSON. Output must match the output format below.
3. Be concise. No preambles, narration, or filler text.

Parse Arguments

Arguments: $ARGUMENTS

• path: Directory or file to scan (default: current working directory)
• --rules <ruleset>: Semgrep ruleset (default: auto)
• --programmatic: Return structured JSON only, with no prose or status messages. Designed for callers like the static analysis pre-pass in /code-review. When set, skip installation guidance on failure — just return the skip status JSON.

Examples:

/semgrep-analyze
/semgrep-analyze src/
/semgrep-analyze --rules p/security-audit
/semgrep-analyze src/utils --rules p/javascript
/semgrep-analyze --programmatic src/

Steps

1. Check Semgrep installation

semgrep --version

If not installed, output:

{"status": "skip", "issues": [], "summary": "semgrep not installed — install via pip install semgrep, pipx install semgrep, or brew install semgrep"}

If --programmatic is set, return the JSON above and stop — do not add installation guidance or prose.

Otherwise (human-facing run), also tell the user: Run /project-init to set up this repo's tooling as the one-stop entry point — it installs semgrep as a capability tool (see its ${CLAUDE_PLUGIN_ROOT}/skills/project-init/references/capability-tools.md) — or install semgrep directly — pip install semgrep (also pipx install semgrep or brew install semgrep). (The direct install command is the fallback.)

Stop.

2. Run Semgrep scan

semgrep scan --config <ruleset> --quiet --json <path>

Default ruleset is auto. Default path is ..

3. Parse results

Map each Semgrep finding to an issue: