The Agent Skills Directory

[Prompt Injection] (SAFE): No malicious instructions or bypass attempts detected in the prompt logic. The instructions are purely functional for data aggregation.
[Data Exposure & Exfiltration] (SAFE): The skill reads from defined project context directories (.context/, .agents/). There are no network operations, and it does not target sensitive system paths like SSH keys or environment secrets.
[Indirect Prompt Injection] (LOW): The skill processes data from local files which could be manipulated by an attacker to influence the AI's behavior.
Ingestion points: .context/project.md, .context/stack.md, .context/architecture.md, .context/conventions.md, and files in .context/specs/.
Boundary markers: Absent; content is interpolated directly into the dashboard template without explicit delimiters or safety instructions.
Capability inventory: No destructive capabilities; the skill explicitly states it is non-destructive and only reads files.
Sanitization: None detected for the ingested file content.
[Command Execution] (SAFE): While the skill mentions 'running' validation, the implementation context and the 'Non-destructive' guarantee indicate this refers to checking existing file states or outputs from other tools rather than arbitrary shell execution.
[Obfuscation] (SAFE): No encoded strings, homoglyphs, or hidden characters were found in the skill file.

onboard