The Agent Skills Directory

[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted HTML content from external websites.
Ingestion points: Content is fetched from a user-defined URL in SKILL.md using curl.
Boundary markers: Not present; the data is processed directly by parsing scripts.
Capability inventory: The skill uses curl and python3 for data retrieval and analysis but does not perform dangerous operations with the ingested data.
Sanitization: Data is processed using standard libraries (re, json); no explicit instruction filtering is implemented.
[COMMAND_EXECUTION]: The skill includes several examples of shell commands to facilitate the audit process.
Evidence: Uses curl to interact with Algolia APIs and python3 -c for local data processing in SKILL.md and references/analytics-api.md.
Context: These commands are used for legitimate diagnostics and data extraction related to search optimization.
[EXTERNAL_DOWNLOADS]: The skill connects to external services to retrieve index data and analytics.
Evidence: Network requests are directed to *.algolia.net and analytics.algolia.com.
Note: These are official endpoints for the Algolia service and are expected for the skill's functionality.

algolia-search-optimizations