using-agent-skills

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill bundle implements an HTTP cache for documentation which ingests content from external URLs.\n
  • Ingestion points: The sdd-cache hooks intercept and process WebFetch tool outputs from external documentation sources.\n
  • Boundary markers: The sdd-cache-pre.sh hook uses ----- BEGIN CACHED CONTENT ----- and ----- END CACHED CONTENT ----- delimiters to separate cached data from agent instructions.\n
  • Capability inventory: The skill set includes file system modification, shell command execution for build/test processes, and network access tools.\n
  • Sanitization: The system relies on the agent's ability to distinguish content within markers and does not perform additional semantic sanitization of the documentation body.\n- [EXTERNAL_DOWNLOADS]: The sdd-cache hooks use curl to fetch HTTP headers from remote servers for cache revalidation. These requests are directed to URLs provided by the agent during documentation lookups.\n- [COMMAND_EXECUTION]: The package uses several local shell scripts (hooks/*.sh) to handle session management and caching. These scripts execute standard command-line utilities such as jq, curl, sed, awk, and perl to process data and maintain the development environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 06:55 AM
Security Audit — agent-trust-hub — using-agent-skills