synthesis-context-lifecycle
Pass
Audited by Gen Agent Trust Hub on May 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides a methodology for the agent to ingest and process data from project files such as CONTEXT.md and REFERENCE.md. This establishes a surface for indirect prompt injection if those files contain untrusted content.
- Ingestion points: The agent reads project-level markdown files including CONTEXT.md, REFERENCE.md, and monthly logs in the sessions/ directory.
- Capability inventory: The instructions authorize the use of shell commands for Git operations (add, commit, push) and file system modifications (create, write, rewrite).
- Boundary markers: The skill suggests using specific templates, phase headers, and line-count budgets (e.g., ≤150 lines) to delineate context types.
- Sanitization: The protocol does not specify explicit escaping or sanitization of the content being moved between files.
- [COMMAND_EXECUTION]: The Archival and Commit Protocols provide explicit instructions for the agent to execute shell commands. Specifically, the agent is directed to use git add, git commit, and git push to manage the project's context infrastructure.
- [DATA_EXFILTRATION]: The Commit Protocol requires the agent to automatically push changes to a remote repository. This involves a network operation that transfers the contents of context files, including session summaries and project reference data, to an external Git host.
Audit Metadata