The Agent Skills Directory

[SAFE]: The skill implements strong credential safety by explicitly instructing the agent never to request API keys or passwords in chat and to avoid passing them as command-line arguments, recommending interactive login or environment variables instead.
[SAFE]: The skill establishes robust guardrails for high-risk operations, including mandatory dry-runs, human approval gates for production changes, and extensive pre-flight checks to verify data materialization and reversibility.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from the user (specifically the 'reason' field) and interpolates it into CLI commands via the --note flag.
Ingestion points: The reason input field defined in the Inputs section of SKILL.md.
Boundary markers: No explicit delimiters or boundary markers are defined for the interpolation of the reason field into the command line.
Capability inventory: The skill performs shell execution of the bruin CLI and writes log records to the local filesystem (.context/ directory).
Sanitization: There is no mention of sanitization or escaping for the free-text reason input before it is passed to the CLI.

pipeline-backfill