Analyze Context

This skill allows the agent to perceive and break down the current state, context, user input, and data into a clear, structured assessment before taking any action.

Steps

1. Perform a fresh read. Do NOT rely on cached context. Read the relevant files or inputs live from disk/user before proceeding.
2. Identify the subject. Determine what entity, system, or problem space is being analyzed.
3. Map the components. Break the input into its logical sub-parts (e.g., architecture layers, data fields, decision factors).
4. Identify constraints & dependencies. Note what is fixed, what is variable, and what other systems are affected.
5. Surface open questions. Flag any ambiguity that would block action and list them explicitly.
6. Produce the assessment. Output a structured summary with: Subject, Components, Constraints, Open Questions, Recommended Next Step.

Output Format

assessment:
  subject: "<what is being analyzed>"
  components:
    - "<component 1>"
    - "<component 2>"
  constraints:
    - "<constraint 1>"
  open_questions:
    - "<question 1>"
  recommended_next_step: "<single clear action>"

Notes

• This skill is always the first step before any planning, execution, or refactoring task.
• If analyzing a file, invoke Zero-Drift Audit Law: read the file fresh before outputting any assessment.
• For documents > 100K tokens, escalate to the notebooklm toolbox.