The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to download and execute external packages including ruv-swarm and claude-flow@alpha. These are third-party dependencies that are executed at runtime during various workflow stages.
[COMMAND_EXECUTION]: The skill performs significant system-level operations via the GitHub CLI (gh) and bash scripts. These include creating issues, generating workflow files, managing releases, and modifying repository structures. Such capabilities grant the skill broad control over the GitHub environment.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) by processing untrusted data to drive AI decision-making.
Ingestion points: Data enters the agent context via gh pr view (files and labels), gh run view (failure logs), and general codebase analysis.
Boundary markers: There are no documented boundary markers or instructions to the LLM to ignore embedded commands within the ingested data.
Capability inventory: The skill possesses powerful capabilities including gh issue create, gh pr comment, gh pr create, and the generation/execution of new GitHub Action workflows.
Sanitization: There is no evidence of sanitization or validation of the ingested strings before they are interpolated into the swarm coordination logic.
[DATA_EXFILTRATION]: Documentation within the skill suggests the use of GitHub Secrets (secrets.API_KEY, secrets.SWARM_CONFIG) in environment variables. While standard for CI/CD, this handling of sensitive credentials requires careful management to ensure they are not exposed to untrusted execution steps.

github-workflow-automation