The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes content from various local files, which could potentially contain malicious instructions designed to influence the agent's implementation or verification steps.
Ingestion points: SKILL.md (Inputs To Read First) specifies reading docs/specs/<feature-id>.md, feature_list.json, PROGRESS.md, and other durable documentation.
Capability inventory: The agent is instructed to write implementation code, update documentation, and execute shell commands (gradlew, pnpm, init.sh) in SKILL.md (Self-Verify) and references/implementation-rules.md.
Boundary markers: There are no explicit delimiters or "ignore embedded instructions" warnings described for the content being processed.
Sanitization: No explicit validation or sanitization routines are mentioned for handling the contents of the ingested specification files.
[COMMAND_EXECUTION]: The skill workflow involves executing shell commands to perform verification and testing tasks.
Evidence: SKILL.md and references/implementation-rules.md specify running ./gradlew testDebugUnitTest, pnpm test:e2e, and the init.sh script.
[DYNAMIC_EXECUTION]: The agent is tasked with creating or modifying scripts that are subsequently executed as part of the repository's gatekeeping process.
Evidence: references/implementation-rules.md (init.sh Rule) instructs the agent to create or update init.sh as an executable gate to run standard checks.

feature-implementer