The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow that processes external requirements (e.g., user stories, feature requests), creating a surface for indirect prompt injection attacks. If a requirement contains adversarial instructions, the agent might be misled during implementation or testing.
Ingestion points: Requirements are ingested in Phase 1 of SKILL.md and Step 1.1 of references/REFERENCE.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are defined for the input data.
Capability inventory: The skill utilizes Task, Edit, and Create tools, and dispatches sub-agents to perform code changes and run commands.
Sanitization: No sanitization or validation of the requirement text is specified.
[DYNAMIC_EXECUTION]: The skill employs a "Skill Discovery" mechanism in SKILL.md and references/REFERENCE.md (Phase 1.5) that dynamically identifies and invokes other specialized skills based on tech stack detection at runtime. This introduces a dependency on the security and integrity of the local environment's skill repository.
[COMMAND_EXECUTION]: The workflow incorporates the execution of build commands, tests, and linters through the Task tool and specialized sub-agents. While intended for development, this provides a pathway for arbitrary command execution if the agent is misled by malicious requirement input.

implementer