The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including source code, Pull Request descriptions, and requirement documents (PRD), as specified in SKILL.md (Steps 2 and 3) and SNAPSHOT.md (Step 3). There are no mandated boundary markers or explicit sanitization procedures for this data, which creates a potential surface for indirect prompt injection where instructions hidden in the analyzed content could influence the agent's behavior. However, SKILL.md Step 6 describes a manual review process where the primary agent validates sub-agent findings against the source content, acting as a defensive filter.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute various shell commands, including git diff, git log, git status, and ripgrep (rg), to gather context for the review. These operations are documented in SKILL.md Step 2 and SNAPSHOT.md Steps 2 and 3. While these are read-only operations typical of a development tool, they represent a subprocess execution surface involving user-influenced parameters.

code-review