Agentic workflow guard auditor
Skill guorunjie/agentic-workflow-guard/.hermes/skills/agentic-workflow-guard-auditor
Static analysis for AI automation workflows. Find prompt-injection paths, overpowered tools, and write-capable agent jobs before they run.
npx -y skills add guorunjie/agentic-workflow-guard --skill agentic-workflow-guard-auditorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use before granting AI automation write access, reviewing agentic GitHub Actions, Bitbucket Pipelines, GitLab CI, Travis CI, Drone CI, TeamCity, Harness CI/CD, Tekton Pipelines, Argo Workflows, AWS CodeBuild, Google Cloud Build, CircleCI, Azure Pipelines, Jenkins, Buildkite, auditing n8n, Dify, Flowise, Langflow, Node-RED, Make, Pipedream, Airflow, browser automation, or checking MCP tool configs for prompt-injection and side-effect risks.
SKILL.md
7.6 KB, as published. Nobody here has run it
Agentic Workflow Guard Auditor
Purpose
Agentic Workflow Guard is a deterministic scanner for AI automation workflows. Use it to find prompt-injection paths, model-output-to-shell sinks, broad write permissions, risky Bitbucket Pipelines, GitLab CI, Travis CI, Drone CI, TeamCity, Harness CI/CD, Tekton Pipelines, Argo Workflows, AWS CodeBuild, Google Cloud Build, CircleCI, Azure Pipelines, Jenkins, and Buildkite agent jobs, risky MCP tools, Dify, Flowise, Langflow, Airflow AI DAGs, browser automation side effects, and low-code AI steps chained into external side effects.
Workflow
For a new repository, run agentic-workflow-guard init . --profile balanced first to scaffold .awg.yml and a release-tagged GitHub Code Scanning workflow.
Run agentic-workflow-guard doctor . after setup or config edits to verify the repository config, schema annotation, Action workflow, rule-pack lock, and baseline.
- Run
agentic-workflow-guard scan . --format markdownfor local review. - Run
agentic-workflow-guard scan . --format sarifwhen the result should feed GitHub Code Scanning. - Use
agentic-workflow-guard scan . --format sarif --output awg.sariffor GitHub Code Scanning uploads. - Use
agentic-workflow-guard schema reportwhen integrating machine-readable JSON reports. - Use
agentic-workflow-guard schema fixwhen integrating structured fix recipe reports. - Use
agentic-workflow-guard schema configwhen integrating.awg.yml,.awg.yaml, or.awg.jsonrepository settings. - Use
agentic-workflow-guard schema rule-packwhen integrating trusted marketplace metadata. - Use
agentic-workflow-guard schema benchmark-corpusandschema benchmark-reportwhen integrating benchmark metadata. - Use
agentic-workflow-guard scan . --profile strictin write-capable or sensitive automation repositories. - Use
agentic-workflow-guard scan . --baseline .awg-baseline.jsonin existing repositories. - Use
awg-ignore AWI001: reasononly for reviewed suppressions with an audit reason. - Review
Suppressed findingsin JSON and Markdown reports before accepting exceptions. - Prioritize high severity AWI001-AWI006 findings before medium AWI007-AWI010 controls.
- Use
agentic-workflow-guard explain <rule-id>for rule-specific risk and remediation. - Use
agentic-workflow-guard fix . --patchfor reviewable GitHub permission, MCP filesystem scope, and CI dry-run diffs. - Use
agentic-workflow-guard fix . --format jsonoragentic-workflow-guard fix . --format json --output awg-fix.jsonfor recipe confidence, automatic/manual modes, approval snippets, next steps, and PR bot artifacts. - Use
agentic-workflow-guard fix . --applyonly for low-risk GitHub permission downgrades, MCP filesystem read-only scoping, and CI dry-run defaults. - Use
agentic-workflow-guard rules search <platform>to find relevant rule metadata. - Use
agentic-workflow-guard rules registry --format jsonto inspect bundled and community rule-pack aliases and checksums. - Use
agentic-workflow-guard rules verify <file>before trusting external rule packs; it checks schema metadata and checksum. - Use
agentic-workflow-guard benchmarkto verify fixture snapshots. - Use
agentic-workflow-guard benchmark --format jsonfor scored benchmark reports with missing and unexpected rules. - Use
agentic-workflow-guard benchmark corpus --format jsonto expose portable safe/vulnerable benchmark metadata. - Use
agentic-workflow-guard mcp resources --format jsonto expose rules, benchmarks, skill instructions, and remediation playbooks. - Use
docs/demos.mdwhen explaining the fastest GitHub Actions, n8n, MCP, browser automation, and benchmark demos. - Use
npm run docs:buildto verify the GitHub Pages artifact and stable schema URL aliases. - Use
npm run smoke:packagebefore releases to install the packed tarball and run the CLI through npx. - Use
npm run release:prepare -- --version 1.0.1 --dry-runbefore releases to preview package version and release-tag changes. - Use
npm run release:status -- --version 1.0.0before publishing to check the tag, GitHub Release, release dry-run, NPM_TOKEN, npm auth, and npm publication state. - Use
npm run release:verify -- --version 1.0.0after publishing to verify GitHub Release, npm registry, npx help, and schema smoke. - Use
npm run release:sync:checkbefore releases to verify generated rule-pack, benchmark corpus, and MCP resource JSON files. - Use
agentic-workflow-guard agents install <target>to install agent context files into another project. - Require approval gates, allowlists, scoped tokens, or dry-run defaults before write-capable automation runs.
Review Checklist
- Treat GitHub issues, pull requests, Bitbucket pull requests, GitLab merge requests, Travis pull requests, Drone pull requests, TeamCity branches, Harness codebase or trigger context, Tekton params, Argo workflow parameters, CodeBuild webhook context, Cloud Build trigger substitutions, CircleCI branches, Azure Pipelines variables, Jenkins change requests, Buildkite branch/message variables, commit messages, webhooks, emails, and form inputs as untrusted prompt input.
- Never pipe raw model output into shell commands, scripts, release steps, deployment commands, or repository writes.
- Downgrade broad GitHub permissions and tool scopes unless a workflow truly needs them.
- Keep secrets out of agent-visible prompts and environment variables.
- Prefer human review for any agent output that can modify code, tickets, deployments, cloud resources, or customer data.
Commands
agentic-workflow-guard init .agentic-workflow-guard doctor .agentic-workflow-guard scan . --format markdownagentic-workflow-guard scan . --format jsonagentic-workflow-guard scan . --format sarifagentic-workflow-guard scan . --format sarif --output awg.sarifagentic-workflow-guard schema reportagentic-workflow-guard schema fixagentic-workflow-guard schema configagentic-workflow-guard schema rule-packagentic-workflow-guard schema benchmark-corpusagentic-workflow-guard schema benchmark-reportagentic-workflow-guard scan . --profile strictagentic-workflow-guard benchmarkagentic-workflow-guard benchmark --format jsonagentic-workflow-guard benchmark corpus --format jsonnpm run smoke:packagenpm run docs:buildnpm run release:prepare -- --version 1.0.1 --dry-runnpm run release:status -- --version 1.0.0npm run release:verify -- --version 1.0.0 --dry-runnpm run release:sync:checkagentic-workflow-guard mcp resources --format jsonagentic-workflow-guard baseline create .agentic-workflow-guard scan . --baseline .awg-baseline.jsonagentic-workflow-guard fix . --patchagentic-workflow-guard fix . --format jsonagentic-workflow-guard fix . --format json --output awg-fix.jsonagentic-workflow-guard fix . --applyagentic-workflow-guard explain AWI001agentic-workflow-guard rules search githubagentic-workflow-guard rules install core .agentic-workflow-guard rules verify .awg/rules/agentic-workflow-guard-core-rules.jsonagentic-workflow-guard agents install claude .agentic-workflow-guard agents