skill-suggester

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill accesses sensitive local files containing prompt history, which often includes code, logs, and private context.
  • Evidence: Path ~/.local/state/opencode/prompt-history*.jsonl is targeted for reading and parsing in SKILL.md.
  • Exposure Risk: The instructions explicitly require the agent to "Include direct quotes from your past prompts as evidence" when presenting recommendations to the user.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it ingests and processes untrusted historical data without clear boundaries.
  • Ingestion points: The skill reads data from ~/.local/state/opencode/prompt-history*.jsonl (found in SKILL.md).
  • Boundary markers: Absent. There are no instructions to use delimiters or ignore embedded commands within the history files.
  • Capability inventory: The skill performs reading, parsing, and summarization of text content.
  • Sanitization: Absent. The skill does not mention filtering or escaping content from past prompts before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 10:58 AM
Security Audit — agent-trust-hub — skill-suggester