The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted external data that could influence its behavior.
Ingestion points: The skill reads the 'Implementer's proposal summary and diff' and the 'Original issue / CI failure / comment' as defined in SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat these inputs as untrusted data or to ignore instructions embedded within them.
Capability inventory: The skill is explicitly authorized to execute 'Project test/lint commands' in the shell, as noted in the 'Inputs' and 'Rules' sections of SKILL.md.
Sanitization: There are no sanitization steps mentioned to validate that the commands being run are strictly safe tests or linting tools.
[COMMAND_EXECUTION]: The skill is designed to interact with the system shell to execute testing and linting tools.
Evidence: SKILL.md contains instructions such as 'You ran tests (or equivalent)' and 'Do not trust implementer's claim that tests passed — run them.' which directs the agent to execute arbitrary project-defined commands.

loop-verifier