The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources, creating a potential surface for indirect prompt injection.
Ingestion points: The tools defined in SKILL.md (firecrawl_developer_search, firecrawl_search, and firecrawl_scrape) retrieve content from public repositories (issues, PRs, READMEs) and general web pages.
Boundary markers: The instructions advise the agent to "Quote the passage, cite the url," which provides a structural guideline for output but does not establish strict delimiters or "ignore instructions" markers for the ingested data itself.
Capability inventory: The skill allows for the retrieval of arbitrary web content which is then returned to the agent context. If the agent has additional tools (e.g., file system access or shell execution), malicious instructions hidden in the retrieved markdown could attempt to trigger those capabilities.
Sanitization: There is no documentation of sanitization or filtering of the retrieved content to remove potential prompt injection patterns before it is presented to the agent.
[DATA_EXFILTRATION]: The skill performs network operations to external endpoints.
Evidence: The skill documentation references https://api.firecrawl.dev/v2/search/developer for search queries.
Context: This domain is a vendor-owned resource associated with the skill's author ('firecrawl') and is consistent with the skill's primary search functionality. The skill also correctly recommends the use of an environment variable ($FIRECRAWL_API_KEY) for secret management.

firecrawl-developer-index