lore
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its core functionality of harvesting insights from agent journals and logs.
- Ingestion points: During the
HARVESTphase, the skill reads full entries from.agents/*.md, Triage postmortems, and Mend remediation logs. - Boundary markers: While the skill employs semantic similarity thresholds (e.g., >=80% for clustering) and confidence scoring, it lacks explicit delimiters or instructions to ignore potential command-like patterns within the harvested data.
- Capability inventory: The skill has the ability to update the central
METAPATTERNS.mdknowledge base and propagate findings viaLORE_INSIGHTandLORE_ALERTsignals to critical agents such as Architect, Darwin, and Nexus, which could lead to cross-agent instruction poisoning. - Sanitization: There is no documented mechanism for sanitizing or escaping the unstructured content extracted from the journals before it is synthesized into patterns or propagated.
Audit Metadata