atona
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project repositories (requirements, issues, codebase) to drive planning and delivery actions, creating a surface for indirect prompt injection.
- Ingestion points: The skill reads planning inputs, requirements, and project knowledge through native readers and specialists (SKILL.md, references/decomposition.md).
- Boundary markers: Utilizes structured HTML artifacts and behavior contracts to delimit intent, although explicit sanitization or delimiters for interpolated data strings are not detailed.
- Capability inventory: Triggers implementation via 'alaga' (code generation) and manages publication through 'seda-pr' (stewardship of PR/MRs).
- Sanitization: Incorporates human-in-the-loop checkpoints via living HTML plans, uses 'atunwo' for code assessment, and includes premortem checks to identify readiness threats.
- [COMMAND_EXECUTION]: The skill guides the agent to execute project-specific shell commands or queries for operational verification.
- Evidence: The 'rollout-readiness.md' reference instructs the agent to 'Use actual queries, commands or observable signals from the project' to verify invariants and post-change observations.
Audit Metadata