The Agent Skills Directory

[SAFE]: The skill is entirely composed of markdown-based templates and instructional content for social media management. No executable scripts, persistence mechanisms, or obfuscated payloads were found.
[DATA_EXFILTRATION]: The skill documents network operations using curl to fetch data from well-known services (Reddit, Hacker News, and Bluesky via Algolia). These operations are intended for social listening and do not target sensitive local files, credentials, or non-whitelisted domains for exfiltration purposes.
[COMMAND_EXECUTION]: The skill provides command-line recipes using curl, jq, and xmllint for data retrieval and analysis. These are standard development tools used for their intended purpose and do not require administrative privileges or involve dangerous command injection patterns.
[PROMPT_INJECTION]: An indirect prompt injection surface is present as the skill ingests content from external social media platforms for engagement triage. 1. Ingestion points: Reddit JSON API, Hacker News API, Bluesky API, and LinkedIn/X feeds. 2. Boundary markers: None explicitly mentioned in the instructions. 3. Capability inventory: Shell execution for data retrieval and LLM drafting of social media replies. 4. Sanitization: None. This surface is inherent to the skill's purpose of social listening and is considered safe in this context.

social