The Agent Skills Directory

[SAFE]: The skill serves as a template and analytical process for estimating task effort. It operates within a restricted environment using only Read, Glob, and Grep tools, which prevents it from performing dangerous actions such as exfiltration or system modification.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it incorporates external task descriptions and project documentation into its analysis context. 1. Ingestion points: Processes the task-description argument and reads project files like CLAUDE.md, design/gdd/ contents, and the project Backlog board. 2. Boundary markers: No specific boundary markers or 'ignore' instructions are used to delimit external data from the skill's logic. 3. Capability inventory: The skill is restricted to read-only tools (Read, Glob, Grep) and cannot execute code, write files, or access the network. 4. Sanitization: No content filtering or sanitization is applied to the ingested data.
[PROMPT_INJECTION]: The skill contains the phrase 'Verdict: COMPLETE' in Phase 4. While this mimics a system-level response, it appears to be a status indicator for the agent's workflow rather than an attempt to override safety filters or misdirect the analyzer.

estimate