GitHub Actions Self-Hosted Risk Audit

Audit GitHub Actions workflows that use self-hosted runners for untrusted trigger and credential-hardening risks.

Security Scan

This skill is internally consistent: it scans local GitHub Actions workflow YAML files for self-hosted-runner risks, requires only bash/python3, does not ask for credentials or perform network access, and its inputs match its stated purpose.

Purpose & Capability

The name/description match the implementation. The script scans workflow YAML files for self-hosted runner patterns and flags trigger/permission/checkout issues. Required binaries (bash, python3) are appropriate and no unrelated credentials or services are requested.

Instruction Scope

Runtime instructions and the included script operate on local files matched by WORKFLOW_GLOB (default .github/workflows/*.y*ml) and only read those files, apply regex checks, and emit reports. There are no instructions to read unrelated system config, exfiltrate data, or call external endpoints. Note: the tool uses regex-based parsing rather than a YAML parser, so it may be brittle and produce false positives/negatives, but that is a functional (not malicious) limitation.

Install Mechanism

No install spec or external downloads — this is instruction-only plus an included shell script. Nothing is pulled from external URLs or written to system locations by an installer.

Credentials

No required environment variables or credentials. Optional env vars control which files to scan and thresholds; they are configuration-only and proportional to the skill's purpose.

Persistence & Privilege

The skill does not request persistent/global privileges (always:false), does not modify other skills or system-wide configs, and does not store credentials. It runs on demand and can be invoked by the user.

Assessment

This skill appears safe to run locally: it only reads workflow YAML files and reports risky patterns, and it does not require API tokens or network access. Before installing/running, review or run the included script on a non-sensitive copy of your repo if you have concerns; note that the tool uses regex-based checks (not a full YAML AST), so expect occasional false positives/negatives. Also be cautious if you override WORKFLOW_GLOB to scan arbitrary paths — the script will read any files you point it at.

Use this skill to flag risky workflow patterns when jobs run on self-hosted GitHub Actions runners.

What this skill does

• Scans workflow YAML files (.github/workflows/*.yml by default)
• Detects workflows that reference self-hosted runners
• Flags high-risk trigger combinations (pull_request_target, pull_request, issue_comment)
• Flags broad/self-hosted-only runner selection (no extra routing labels)
• Flags workflows with write-capable permissions in self-hosted contexts
• Flags actions/checkout steps that do not set persist-credentials: false
• Supports text/json output and CI fail gate

Inputs

Optional:

• WORKFLOW_GLOB (default: .github/workflows/*.y*ml)
• TOP_N (default: 20)
• OUTPUT_FORMAT (text or json, default: text)
• WARN_SCORE (default: 4)
• CRITICAL_SCORE (default: 8)
• WORKFLOW_FILE_MATCH / WORKFLOW_FILE_EXCLUDE (regex, optional)
• FAIL_ON_CRITICAL (0 or 1, default: 0)

Run

Text report:

WORKFLOW_GLOB='.github/workflows/*.yml' \
WARN_SCORE=4 \
CRITICAL_SCORE=8 \
bash skills/github-actions-self-hosted-risk-audit/scripts/self-hosted-risk-audit.sh

JSON output + fail gate:

WORKFLOW_GLOB='.github/workflows/*.y*ml' \
OUTPUT_FORMAT=json \
FAIL_ON_CRITICAL=1 \
bash skills/github-actions-self-hosted-risk-audit/scripts/self-hosted-risk-audit.sh

Run against bundled fixtures:

WORKFLOW_GLOB='skills/github-actions-self-hosted-risk-audit/fixtures/*.y*ml' \
bash skills/github-actions-self-hosted-risk-audit/scripts/self-hosted-risk-audit.sh

Output contract

• Exit 0 in reporting mode (default)
• Exit 1 when FAIL_ON_CRITICAL=1 and one or more workflows are critical
• Text mode prints summary + top flagged workflows
• JSON mode prints summary + flagged workflows + critical workflows