oodle-o11y-review
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted code diffs during the PR review or authoring process. This creates a vulnerability surface where a malicious actor could embed instructions within code comments or string literals intended to deceive the agent or bypass its review logic.
- Ingestion points: The agent reads the diff of a code change, which constitutes untrusted external data (SKILL.md).
- Boundary markers: The instructions lack specific delimitation or guidance to treat code content as data rather than instructions, potentially leading to accidental obedience of embedded commands.
- Capability inventory: The skill has the capability to write findings to a manifest file ('O11y Change Context') and interact with observability tools via MCP, providing a path for influence if the agent is compromised by injected content.
- Sanitization: No sanitization or filtering of the input code diff is specified before the agent processes it.
- [DYNAMIC_EXECUTION]: The skill uses a dynamic discovery process to identify observability backends and load corresponding workflow skills (referenced as 'load_skill__*'). This involves runtime selection of executable skills based on environment characteristics (SKILL.md).
- Evidence: The 'Backend discovery' section instructs the agent to 'load a matching load_skill__* when one fits' based on the functional shape of the tools detected in the environment.
Audit Metadata