data-management-plan

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data such as research specifications, grant drafts, and interview outputs provided via the --input parameter.
  • Ingestion points: Research specs and draft documents are read from the file system to extract data types and identification strategies.
  • Boundary markers: The skill uses a [CLARIFY: <specific question>] placeholder convention to handle missing or ambiguous information, reducing the likelihood of the model following instructions embedded within the input data.
  • Capability inventory: The skill uses Write to save the draft DMP to the local file system and Task to perform citation verification.
  • Sanitization: The output is a markdown draft intended for human review and manual pasting into official funder portals, providing a mandatory human-in-the-loop checkpoint.
  • [COMMAND_EXECUTION]: The skill uses the Task tool in Phase 4 to invoke internal verification mechanisms for funder policy citations. These are platform-internal operations directed at specialized sub-skills (/verify-claims, /capture-environment, /replication-package) and do not involve arbitrary shell command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 06:31 PM
Security Audit — agent-trust-hub — data-management-plan