The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill architecture allows for the processing of untrusted data which is subsequently fed into the execution context of sub-agents.
Ingestion points: The skill reads output from git log (commit messages), git diff (code changes), and the contents of external specification files located in directories like .scratch/, docs/, or specs/.
Boundary markers: There are no explicit delimiters or "ignore embedded instructions" warnings used when passing these external inputs into the sub-agent prompts defined in the "Spawn both sub-agents in parallel" step.
Capability inventory: The skill possesses the ability to execute shell commands (via git) and perform file system writes to update local markdown files.
Sanitization: The instructions do not prescribe any escaping, validation, or filtering of the external content before it is interpolated into the prompts for the Agent or Task tools.
[COMMAND_EXECUTION]: The skill directs the agent to execute multiple git commands (git diff, git log, git rev-parse) using a "fixed point" provided by the user. If the agent's execution environment does not properly sanitize this input, it could lead to command injection.
[DYNAMIC_EXECUTION]: The skill instructions define a process where prompts for sub-agents are dynamically constructed via string concatenation of command outputs and file contents. This generated prompt content is then executed using the platform's sub-agent invocation tools (Agent in Claude Code or Task in Cursor).
[EXTERNAL_DOWNLOADS]: The skill includes functionality to fetch data from remote issue trackers such as GitHub and GitLab to identify the technical specifications for the code review, depending on the repository configuration.

review-axes