The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions and shell script templates for the agent to execute system commands. These include the anycap CLI for data retrieval, as well as jq, sed, and column for parsing JSON output and generating Markdown tables. These operations are restricted to the local environment and intended for data processing.
[EXTERNAL_DOWNLOADS]: The skill fetches data from external web sources using the anycap search and anycap crawl functions. It specifically highlights Wikipedia as a reliable source for verifying player information. This behavior is documented as the core functionality of the skill.
[PROMPT_INJECTION]: As the skill ingests and processes untrusted data from web search results to generate predictions, it contains a surface for indirect prompt injection. The skill mitigates this by requiring a mandatory entertainment-only disclaimer.
[PROMPT_INJECTION]: Indirect Injection Analysis: 1. Ingestion points: Output from anycap search and crawl operations (SKILL.md, verify.md). 2. Boundary markers: No explicit delimiters provided for external content processing. 3. Capability inventory: Execution of anycap, jq, and sed; local file writing to /tmp/ and squad.json (batch.md, verify.md). 4. Sanitization: Structural validation via jq parsing of CLI results.
[SAFE]: All identified behaviors are consistent with the skill's stated purpose of providing football match predictions and intelligence. The use of the anycap-ai vendor's own tools and reputable sources like Wikipedia follows expected usage patterns for this agent extension.

anycap-worldcup-predict