addy-orchestrator
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill maintains persistent state by reading from and writing to a local file named '.addy-session.md' in the project directory, which is necessary for tracking workflow progress.
- [SAFE]: The orchestrator identifies and utilizes sub-skills from a predefined local plugin cache path (~/.claude/plugins/cache/addy-agent-skills/agent-skills/1.0.0/skills/) to execute specific development phases.
- [PROMPT_INJECTION]: The skill ingests data from the project state file and interpolates it into prompts for sub-agents, creating an indirect prompt injection surface.
- Ingestion points: Reads previous workflow state and context from the project-level .addy-session.md file.
- Boundary markers: Employs structural delimiters such as '== Accumulated context ==' to separate historical data from phase instructions.
- Capability inventory: Manages file read/write operations and coordinates the execution of various developer skills.
- Sanitization: The skill does not perform explicit validation or sanitization of the context block before passing it to subsequent agents.
Audit Metadata