The Agent Skills Directory

Behavioral Directives: The skill employs strong instructions and emotional framing to enforce a specific persona and prevent the omission of workflow steps. This includes persistent requirements to load the skill and directives to push back against user input, which serve to override default agent tendencies.
Indirect Instruction Surface: The agent is directed to ingest information from subagent research and external skill lists. 1. Ingestion points: SKILL.md (Subagent research results and skill search). 2. Boundary markers: Absent; there are no specified delimiters to isolate external research data from the system prompt. 3. Capability inventory: SKILL.md (TodoWrite tool usage, git environment checks, and branch management). 4. Sanitization: Absent; the skill does not detail methods for sanitizing the outputs of research subagents before processing.
System Interaction Patterns: The workflow involves executing local environment checks like git status and managing git branches. While typical for development assistants, these represent execution of shell commands within the local environment.

starting-nori