The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including issue reports, wiki pages, and roadmap goals to synthesize a "problem map" and guide decision-making. Evidence: The gathering process is described in SKILL.md (Flow step 4) and references/problem-altitudes.md. While the skill includes internal reasoning steps, it lacks strict boundary delimiters for untrusted evidence content, creating a surface where malicious instructions in reports could influence the agent's classification or downstream routing to implementation skills like /cook.
[COMMAND_EXECUTION]: The skill invokes platform commands and other skills including /wheypoint, /mold, /cook, and /briesearch as part of its handoff process. Evidence: Handoff logic is defined in SKILL.md under ## Handoff slug and ## Handoff.

culture