The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a significant surface for indirect prompt injection as its primary function is to ingest and analyze untrusted data from multiple external and internal sources.
Ingestion points: The skill reads local project files (README, specifications, source code), the output of gh issue list, and external content fetched via scripts/evidence.py from the App Store, Hacker News, GitHub, and Stack Overflow.
Boundary markers: The instructions in SKILL.md do not implement explicit delimiters or instruction-ignoring warnings when the agent processes the ingested data, increasing the risk that embedded malicious instructions could influence the agent's output.
Capability inventory: The agent has the ability to execute grep for file searching, the gh CLI for repository metadata, and a custom Python script for network requests. Its primary output is the generation of a markdown report.
Sanitization: While scripts/evidence.py performs basic text cleaning for display purposes (HTML unescaping), it does not provide security-focused sanitization to prevent prompt injection attacks originating from the external sources it queries.
[EXTERNAL_DOWNLOADS]: The skill fetches data from several well-known external services to gather evidence for its product audits.
Evidence: The scripts/evidence.py tool connects to official APIs for Apple's App Store, Algolia (Hacker News), GitHub, and StackExchange.
Context: These network operations are necessary for the skill's intended functionality and target established, reputable service providers.
[COMMAND_EXECUTION]: The skill's workflow involves executing shell commands to inspect the project and gather data.
Evidence: SKILL.md instructs the agent to use grep for searching its internal knowledge base, gh issue list to read project feedback, and python3 to run the provided evidence collection script.
Context: These commands are used as intended for local project analysis and knowledge retrieval.

product-thinking