The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's code-reviewer subagent processes external data that could contain hidden instructions to override the agent's behavior.
Ingestion points: External data enters the subagent's context through the {WHAT_WAS_IMPLEMENTED}, {PLAN_OR_REQUIREMENTS}, and {DESCRIPTION} placeholders in code-reviewer.md, as well as via the output of git diff which inspects the project's source code.
Boundary markers: Absent. The code-reviewer.md template interpolates these variables directly into the prompt without using specific delimiters or warnings to ignore instructions embedded within that content.
Capability inventory: The subagent is instructed to execute git diff (in code-reviewer.md), and the primary agent executes git rev-parse and git log (in SKILL.md).
Sanitization: Absent. There is no evidence of escaping, validation, or filtering of the content retrieved from the plan files or the git repository before it is processed by the LLM.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands to interact with the local filesystem and version control system.
Evidence: SKILL.md contains instructions to run git rev-parse and git log to identify commit ranges. code-reviewer.md instructs the subagent to run git diff --stat and git diff to analyze code changes.

requesting-code-review