The Agent Skills Directory

[SAFE]: The skill performs legitimate performance analysis on CI logs from the official withastro/astro repository.
[SAFE]: All external resources and repositories belong to the verified vendor (withastro).
[SAFE]: Command execution is limited to standard GitHub CLI operations and the provided parsing script, with no evidence of malicious intent or obfuscation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external CI logs, representing a theoretical vulnerability surface.
Ingestion points: CI logs downloaded via gh run view as instructed in SKILL.md.
Boundary markers: None explicitly defined in the agent instructions.
Capability inventory: The skill uses gh for log retrieval and node to execute a local script; no arbitrary network or high-privilege operations are present.
Sanitization: The parse-test-durations.js script uses specific regular expressions to filter log lines but does not sanitize the content of extracted test names. This is a low-risk surface common to log-parsing utilities.

astro-test-perf