The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests outputs from conflicting agents and research tracks to generate new skill templates. While this creates a surface for potentially malicious instructions to be included in generated drafts, the process is mitigated by a mandatory human confirmation gate ('I will place this draft at {path}. Shall I proceed?') before any files are written to the system. Ingestion points: conflict data from skills/agents/research tracks (Step 1). Boundary markers: absent in prompt interpolation. Capability inventory: uses the Write tool to generate SKILL.md skeletons (Step 4). Sanitization: absent.- [COMMAND_EXECUTION]: The skill utilizes Bash and Grep for analyzing conflict points within research tracks and skill outputs. This behavior is consistent with the skill's primary purpose of text analysis and metadata harvesting.

contention-layer