skills/yogsoth-ai/de-anthropocentric-research-engine/audit-explanatory-compression/Gen Agent Trust Hub
audit-explanatory-compression
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill is entirely composed of YAML-formatted logic and Markdown instructions. It defines high-level protocols for an agent to audit explanations for compression efficiency.
- [INDIRECT_PROMPT_INJECTION]: The skill defines an input surface for processing external data (explanations, covered facts, and alternatives). While this represents a theoretical attack surface for indirect prompt injection, the skill does not request any dangerous capabilities such as shell execution, network access, or file system modifications, rendering the risk negligible.
- Ingestion points: Ingestion is defined via the
mode_contractsinSKILL.md(fields:explanation,covered_facts,alternatives). - Boundary markers: The instructions do not specify explicit delimiters or "ignore embedded instructions" warnings for the ingested data.
- Capability inventory: No tools or subprocess calls are requested in the skill's frontmatter or body.
- Sanitization: No explicit sanitization or validation logic is defined for the input fields.
Audit Metadata