The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the repository being worked on.
Ingestion points: Files read in Step 1 (Understaning the Task), Step 3 (Reference Implementation globbing), and Step 4 (Reading files before editing).
Boundary markers: Absent. The instructions do not define delimiters to separate file content from the agent's instructions.
Capability inventory: Execution of shell commands via pnpm, file modification via Write/Edit tools, and the ability to invoke other domain-specific skills.
Sanitization: Absent. The skill processes the content of read files directly without validation or filtering.
[COMMAND_EXECUTION]: The skill executes shell commands as part of its verification loop.
Evidence: In Step 5, the skill runs pnpm test and pnpm run typecheck to verify code changes. In Step 6, it runs git diff to confirm the scope of modifications. These are standard operations for a development skill.

dev