The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze output from external network targets (e.g., HTTP responses and LDAP search results) to identify vulnerabilities. A malicious target could return crafted content or error messages designed to influence the agent's logic or bypass its instructions.
Ingestion points: Output of curl and ldapsearch commands in SKILL.md used for probing and response analysis.
Boundary markers: Absent; the skill does not define specific delimiters to wrap external content or include warnings to ignore instructions embedded in target responses.
Capability inventory: Includes curl for HTTP network requests, ldapsearch for LDAP protocol interaction, grep for pattern matching, and cat for reading local files (e.g., scope.txt).
Sanitization: No explicit sanitization or filtering of the ingested external data is described; the analysis focuses on identifying specific vulnerability signals in the raw output.
[COMMAND_EXECUTION]: The skill provides executable command templates that perform network operations and probe remote systems.
Evidence: Use of curl -s -X POST https://$TARGET/api/login for vulnerability testing in SKILL.md.
Evidence: Use of ldapsearch -W to interact with directory services in SKILL.md.

hunt-ldap