The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform local testing activities, including running unit and integration tests as well as smoke checks for systems affected by the patch. These operations are conducted within the local project environment to verify code stability.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface through the ingestion of external bug reports.
Ingestion points: Untrusted data is read from production/qa/bugs/*.md in Phase 1 and Phase 4.
Boundary markers: The skill does not explicitly utilize boundary markers or delimiters when passing bug report content to the lead-programmer or qa-tester sub-agents.
Capability inventory: The skill and its sub-agents have access to sensitive tools such as Edit, Write, and Bash, which could potentially be misused if malicious instructions are embedded in the bug data.
Sanitization: There is no mention of sanitizing or validating the contents of bug reports before they are provided to the agentic implementation loop.

day-one-patch