The Agent Skills Directory

[SAFE]: The skill is entirely instructional and does not contain any executable code, shell scripts, or binaries. Its operations are limited to text analysis, multi-agent reasoning, and documenting results in a localized directory (docs/ideas/).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied product ideas which could theoretically contain malicious instructions. However, the 'Independence Protocol' defined in SKILL.md mitigates this risk by requiring the agent to isolate analytical lenses, categorize inputs as fact/inference/assumption, and synthesis findings only after independent review. This structured workflow provides strong protection against accidental obedience to instructions embedded in data.
[EXTERNAL_DOWNLOADS]: The skill references academic research on arXiv and a reference skill definition in a public GitHub repository (obra/superpowers). These are static references used for methodological grounding and documentation. They do not involve runtime downloading or execution of scripts.
[DATA_EXPOSURE_EXFILTRATION]: The skill defines an output contract to save results to the filesystem. This is a standard and safe practice for an AI agent performing documentation and decision-tracking tasks, focused on the current project's context.

hotseat