The Agent Skills Directory

[COMMAND_EXECUTION]: The skill parses $ARGUMENTS and passes them to command-line tools. Without strict sanitization, this pattern allows for command injection vulnerabilities where a user or malicious prompt could execute arbitrary shell commands.\n- [DYNAMIC_EXECUTION]: The workflow explicitly runs tests and harnesses from the target repository during the 'runtime' and 'bug-fix-verifier' stages. This involves executing arbitrary code defined in the repository being analyzed, which is a significant security risk if the repository is untrusted.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes bug reports and repository content, which are untrusted external inputs. These inputs are passed to triage agents and verifiers, creating a surface for indirect prompt injection.\n
Ingestion points: [BUG] issue descriptions and repository code files (SKILL.md)\n
Boundary markers: Mentions use of 'ingest boundaries' and 'batch paths' for data isolation.\n
Capability inventory: Employs Bash and Task tools for executing shell commands and tests.\n
Sanitization: No explicit sanitization or filtering logic is detailed for the external bug report content.

analyze-bugs