The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository's CHOICES.md file to determine the scope, metric, and gate for its automated improvement cycles.
Ingestion points: The skill parses parameters from CHOICES.md in Step 1 of the loop.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands within the parsed parameters are mentioned.
Capability inventory: The skill possesses the ability to propose Pull Requests (Step 3), execute arbitrary code modifications via the /task skill (Step 5), and write data to the objective_metrics file (Step 2 and 6).
Sanitization: There is no evidence of sanitization or validation of the input parameters parsed from the repository files before they are used to drive the improvement loop.

hillclimb