omh-paper-learning

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external paper content (PDFs, arXiv entries, text excerpts), creating a data ingestion surface. 1. Ingestion points: User-provided attachments and text excerpts described in example prompts. 2. Boundary markers: The skill instructs the agent to maintain a 'coverage ledger' and record 'source_state' (e.g., metadata_only, excerpt_text_observed) to delineate verified from unverified content. 3. Capability inventory: The agent can execute omh CLI commands for state management and write to the $OMH_HOME directory. 4. Sanitization: The skill emphasizes evidence validation and labeling (e.g., 'excerpt explanation') but does not specify input sanitization.
  • [COMMAND_EXECUTION]: The skill instructions utilize the omh CLI toolset (omh paper plan, omh paper progress, omh runtime record) to document the research process. These commands manage state transitions and append to progress logs in the local $OMH_HOME/paper-learning/ directory. The usage is consistent with workflow management and does not involve high-risk patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 08:34 PM
Security Audit — agent-trust-hub — omh-paper-learning