competitive-product-research

This skill appears internally consistent and low-risk, but check these practical points before installing or running it: - Verify browsing/data access: the SKILL.md expects the agent to gather public information (URLs). If your agent instance lacks safe web browsing/tools, results may be low-quality or the model may hallucinate sources — always check SRC entries the skill generates. - Do not supply secrets: the skill only needs public info and optional screenshots/docs; never provide credentials, keys, or internal secrets when asking it to collect data. - Validate sources and findings: the skill mandates source citation, but models can still fabricate — spot-check that cited URLs and SRC items actually support the claims before acting on recommendations. - Privacy of user materials: uploaded screenshots/recordings may contain internal info; confirm you are permitted to share them with the skill and that the 'shareable' output mode will properly redact sensitive fields. - Upstream provenance: the package metadata (claw.json) lists a GitHub repo and a support URL — if you want stronger assurance, review that repository and the author (Chris1Wang3) before widespread use. If you need the skill to fetch live web data, ensure your agent platform enables a controlled browsing tool or provide vetted source URLs yourself to reduce hallucination risk.