The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions as an orchestrator that synthesizes information from user task descriptions and multiple simulated sub-agent personas to generate execution plans. This architecture creates an indirect prompt injection surface where instructions hidden in the task input or sub-agent outputs could influence the final plan.
Ingestion points: Processes user-provided task descriptions and hypothetical outputs from specialized sub-agents (e.g., '007', 'skill-sentinel') mentioned in SKILL.md.
Boundary markers: The planning templates (Fase 4, 5, and 6) lack explicit delimiters or instructions to ignore embedded commands within interpolated content.
Capability inventory: The outlined workflows involve sensitive actions such as file system modifications, environment configuration writes, and version control operations (git status/commit).
Sanitization: The skill provides defensive guidance, such as requiring YAML schema validation for metadata and SDK-based signature verification (HMAC-SHA256) for webhooks.
[EXTERNAL_DOWNLOADS]: The documentation in references/problem-catalog.md mentions the use of pip install -r requirements.txt as a standard solution for resolving missing dependencies.

task-intelligence