The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Hardcoded Apify user ID (wjz2NfU1Y0MdMxeey) discovered in SKILL.md.
[DATA_EXFILTRATION]: The interactive rating page (https://superseller.agency/compete/[tenantSlug]) is explicitly documented as having "no auth required". This allows unauthorized parties to access collected competitor research, AI-generated replication briefs, and potentially sensitive client feedback if they can guess or obtain the tenant slug.
[COMMAND_EXECUTION]: The workflow for adding new customers involves cloning and running scripts using shell commands with a placeholder <customer> variable (cp ... <customer>-... and npx tsx ...). This pattern is highly susceptible to command injection if the customer slug is not strictly validated before being used in the shell context.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted data from the Meta Ads Library and Instagram/Facebook follower bios.
Ingestion points: Meta ad copy/titles (ad_text, ad_title) and social media bios (profile_bio).
Boundary markers: No explicit delimiters or instruction-ignore warnings are used when passing this data to Claude for extraction and briefing.
Capability inventory: The skill has database write access (PostgreSQL), shell execution capabilities via tsx, and network access via Apify actors.
Sanitization: No sanitization or filtering of the external content is mentioned before it is processed by the AI models.

competitor-research