memory-and-handoff
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manages durable session state within a .cc10x/ directory and generates portable handoff artifacts. These operations utilize standard filesystem tools (Read, Write, Edit, Bash) for legitimate project maintenance and documentation purposes.
- [SAFE]: The skill includes an explicit 'Secret Redaction' policy that mandates replacing tokens, API keys, and PII with redaction markers before persisting content to outward-facing artifacts. This proactive approach significantly reduces the risk of credential leakage during context sharing.
- [COMMAND_EXECUTION]: The skill uses the Bash tool for administrative tasks such as creating directories (e.g., mkdir -p .cc10x) and executing local verification tests (e.g., npm test). These commands are localized to the project environment and do not involve remote execution or privilege escalation.
- [PROMPT_INJECTION]: The skill ingests data from memory files and project documentation, representing a surface for indirect prompt injection. This risk is addressed via a 'Compaction KEEP / SUMMARIZE / DROP' rubric that prioritizes distillation over verbatim ingestion. * Ingestion points: Memory markdown files (.cc10x/.md) and project documentation (docs/plans/, docs/research/*). * Boundary markers: Data is segmented using stable markdown anchors and specific protocol-defined headings (e.g., ## Learnings, ## Decisions). * Capability inventory: The skill utilizes Bash, Write, and Edit tools for local filesystem and documentation management. * Sanitization: Implements a mandatory secret redaction process and a distillation rubric designed to filter out unnecessary or potentially malicious prose.
Audit Metadata