The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines an Iterations: unlimited flag, allowing for unbounded autonomous loops. This creates a risk of resource exhaustion or the agent becoming trapped in infinite cycles if termination criteria are not reached.
[INDIRECT_PROMPT_INJECTION]: The skill utilizes an autonomous modify → verify loop that processes content from external files like handoff.json, result logs, and the target codebase. This architecture creates a vulnerability surface where malicious data could influence the agent's logic. Ingestion points: Untrusted data is ingested via handoff.json, *-results.tsv, and codebase files cited by the analysis personas. Boundary markers: No explicit delimiters or boundary markers are defined to separate instructions from untrusted data in the iteration context. Capability inventory: The skill performs file system modifications, executes verification tools, and manages deployment phases. Sanitization: The protocol lacks evidence of input validation or sanitization for external content before it is interpolated into the agent's prompts.

autoresearch