melech-consult
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it is designed to ingest and process untrusted data from repository artifacts (code, diffs, docs, logs) and interpolate them into prompts for fresh model consultants.
- Ingestion points: The workflow in
SKILL.md(Steps 2 and 3) explicitly instructs the agent to read file paths, logs, and data to build a "neutral packet" for analysis. - Boundary markers: The skill implements a robust logical boundary by mandating that consultants receive only the raw artifact and never the conversation history or the author agent's narration, which significantly reduces the risk of sub-agents being influenced by embedded instructions or previous context.
- Capability inventory: The skill possesses the capability to perform file system reads via tool calls and dispatch concurrent model requests to different providers.
- Sanitization: The instructions do not specify automated sanitization or escaping of the ingested artifact content, relying instead on the isolation of the consultant prompts to mitigate risks.
Audit Metadata