kingkongshot/pensieve

Overview

This skill routes user intent to the correct Pensieve tool or pipeline and enforces the plugin's operational rules. It loads immediately whenever a user expresses intent and guards user data stored at the project-level .claude/pensieve/ directory. The skill prioritizes explicit commands, version checks, and strict tool contract enforcement before any tool runs.

How this skill works

On each user input the skill first checks for explicit commands (loop, /selfimprove, /upgrade, /init, pipeline, doctor) and routes accordingly. If no explicit command is present it infers a candidate intent from the session stage and proposes an action, always asking the user to confirm before executing. Before any tool run it performs the version pre-check and enforces the tool contract: validate 'Use when', required inputs, output contract, and failure fallback. User data lives in .claude/pensieve/ and is never overwritten by the skill.

When to use it

• Whenever you express intent — the skill loads automatically to route you.
• Say explicit commands like loop, init, upgrade, pipeline, doctor, or /selfimprove to run that tool directly.
• When you want to initialize project storage and seed files for Pensieve.
• When a complex task should run as an automated loop or needs repeated split-and-execute steps.
• When you want to improve Pensieve behavior or pipelines (use Self-Improve).
• When you suspect version mismatches, compatibility problems, or need migrations (run upgrade first).

Best practices

• Always issue explicit commands for high-priority actions (upgrade, loop, init, self-improve, doctor, pipeline).
• If you don’t issue a command, answer the single-line confirmation question to avoid accidental execution.
• Mention version or compatibility concerns immediately so the skill can run the upgrade pre-check first.
• Keep project-level user data under .claude/pensieve/ and add decisions, maxims, knowledge, pipelines, and loop outputs there.
• When requesting changes to Pensieve behavior, request Self-Improve so the skill applies the proper improvement workflow.

Example use cases

• Start a new project: say init to seed .claude/pensieve/ with standard files and pipelines.
• Run a complex multi-step task: say loop to split the task and run automated loops.
• Improve pipelines or rules: ask to self-improve and the skill will route you to the Self-Improve tool.
• Check health or validate formats after an upgrade: run doctor for structural and format checks.
• Resolve version concerns or migrate legacy data: say upgrade to run the version pre-check and migrations.