The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a local bootstrap script scripts/bootstrap-experience.ts to initialize the .agents/experience/ directory structure. The script is designed to create append-only markdown files for storing project context. It includes security checks using lstatSync to ensure it does not operate on symbolic links, preventing potential symlink-based attacks.
[EXTERNAL_DOWNLOADS]: The baseline-collector-agent is tasked with performing web searches to retrieve industry benchmarks (e.g., SaaS conversion rates) when user-specific metrics are not available. This retrieval of external data is essential for the skill's purpose of setting realistic business targets.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the user brief and upstream artifacts like prioritize.md (referenced in SKILL.md). The ingestion surface is managed through a multi-agent pipeline where the critic-agent and sanity-check-agent (defined in agents/critic-agent.md and agents/sanity-check-agent.md) verify that the generated outputs meet specific quality and safety gates, such as checking for vanity metrics or aspirational math without justification.

funnel-planner