The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes CI logs fetched from the withastro/astro repository. This content is untrusted data that could contain malicious instructions designed to override agent behavior when the final report is generated.
Ingestion points: The skill downloads logs to /tmp/ci-run-<run_id>.log and reads them via scripts/parse-test-durations.js.
Boundary markers: The logs are converted to JSON, but the skill lacks explicit boundary markers or instructions to the agent to disregard instructions embedded within test names or output content.
Capability inventory: The skill uses gh CLI for network/repository access and executes a local Node.js script to process files.
Sanitization: While the script cleans ANSI escape codes, it does not sanitize or filter the extracted text for prompt injection patterns before inclusion in the final report.
[COMMAND_EXECUTION]: The skill relies on executing the gh (GitHub) CLI tool to list, view, and download CI logs, and uses node to execute a bundled parsing script.
[EXTERNAL_DOWNLOADS]: The skill fetches CI run data and logs from GitHub via the gh CLI. These resources originate from the vendor's own repository (withastro/astro).

astro-test-perf