The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to interact with the local git environment and run verification tools.\n
Evidence: SKILL.md utilizes git rev-parse, git diff, and git ls-files to identify the files to be reviewed.\n
Evidence: The Verification section in SKILL.md describes running formatters, linters, and type-checkers on files modified in the session.\n- [REMOTE_CODE_EXECUTION]: The verification process involves executing test suites, which can run arbitrary code from the repository being reviewed.\n
Evidence: SKILL.md (Verification section) directs the agent to run "targeted tests for impacted modules." If an attacker submits a malicious Pull Request containing compromised test logic, the agent will execute that code on the host system during the verification step.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted external code.\n
Ingestion points: The skill reads code diffs and file contents as part of its primary workflow (Step 1 in SKILL.md).\n
Boundary markers: There are no explicit markers or instructions defined to prevent the agent from following instructions embedded within the reviewed code's comments or data.\n
Capability inventory: The skill has the capability to modify files (via the --fix argument) and execute shell commands (Verification steps).\n
Sanitization: No sanitization or validation of the input content is performed to filter out potentially malicious instructions.

code-review