The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to run the tsx TypeScript executor. This utility downloads the required package from the npm registry if it is not cached locally.
[COMMAND_EXECUTION]: The skill is configured to use several shell commands via Bash, including docker, gh (GitHub CLI), npx, npm, and grep. These are used to fingerprint the execution environment and search through diagnostic records.
[DYNAMIC_EXECUTION]: Local TypeScript scripts (scripts/diagnostics/cli.ts) are executed at runtime using the tsx runner. This is a standard method for executing development scripts but constitutes dynamic code execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as error strings and workflow logs provided by the user to diagnose failures.
Ingestion points: Error strings and workflow run symptoms supplied by the user (as referenced in SKILL.md) are used as search queries in diagnostic scripts.
Boundary markers: The instructions recommend using "redacted error strings" to minimize risk.
Capability inventory: The skill can perform network operations via gh and npm, and read local files (specified in allowed-tools).
Sanitization: Explicit safety rules prohibit the AI from requesting API keys, tokens, JWTs, or environment dumps, significantly reducing the impact of potential injection attacks.

diagnose-awf