thinking

Plugin thinking from marketplace turtlestack in hpsgd/turtlestack. Commands read from the repo's manifest — 1 month ago.

Skills (16)

• algorithm

  Structured seven-phase execution methodology for complex tasks. Transition from current state to ideal state using verifiable criteria. Use for any non-trivial work that benefits from systematic execution.

• council

  Simulate a structured debate between diverse expert perspectives. Use when weighing options, making decisions, or needing multiple viewpoints on a problem.

• creative

  Divergent ideation and creative brainstorming. Use when you need novel solutions, want to break out of conventional thinking, or need multiple distinct approaches.

• first-principles

  Deconstruct a problem to fundamental truths and rebuild from scratch. Use when stuck, challenging inherited constraints, or making architecture decisions where convention may be wrong.

• handoff

  Write or resume a session handoff doc. Use when ending a session with in-flight work that another session will pick up, or starting a session to continue prior work. Captures state explicitly so the next session doesn't have to reconstruct it from chat scrollback.

• health-check

  Generate a health report of the current project's Claude Code setup — installed rules, enabled plugins, memory state, wisdom frame health, and learning coverage.

• isc

  Decompose a request into Identifiable, Specific, Verifiable Criteria before executing. Use at the start of any non-trivial task to ensure nothing is missed.

• iterative-depth

  Analyse a problem through multiple structured passes from different lenses. Use for requirements analysis, architecture decisions, or any problem that benefits from multi-angle examination.

• learning

  Capture, categorise, and recall learnings from work sessions. Use after completing work, when something unexpected happens, or when explicitly asked to remember something.

• propose-improvement

  Propose a change to a marketplace repo based on learned patterns — new rules, updated skills, evolved regex patterns. Infers which upstream marketplace the learning belongs to, confirms with the user, then creates a branch, applies changes, shows diff for review, and raises a PR on approval. Use whe

• reconcile-rules

  Review learned rules against marketplace rules for overlap. Identifies learned rules that have been superseded by upstream changes and recommends cleanup. Use periodically to reduce context bloat from duplicate rules.

• red-team

  Adversarial analysis to stress-test an idea, plan, or argument. Use to find every weakness before they find you. Purely adversarial — for balanced debate, use /council instead.

• retrospective

  Analyse a session transcript for learnings — corrections, reversals, approach changes, and successes. Runs automatically on SessionStart for the previous session. Invoke manually to analyse the current session or review accumulated learnings.

• review-settings

  Audit Claude Code settings.json files for overly broad permissions, overlapping rules, and cross-file redundancy. Produces a reconciled config.

• scientific-method

  Apply the scientific method to any problem — define goal, hypothesise, experiment, measure, iterate. The meta-skill governing all structured investigation. Use when debugging, validating assumptions, or turning uncertainty into knowledge.

• wisdom

  Build and query domain-specific wisdom frames — crystallised patterns from accumulated experience. Use to capture high-confidence principles or to consult accumulated wisdom before making decisions.

Hooks (5)

• SessionStart

  bash ${CLAUDE_PLUGIN_ROOT}/scripts/install-rules.sh

• SessionStart

  bash ${CLAUDE_PLUGIN_ROOT}/scripts/learning-hook.sh "${CLAUDE_PLUGIN_ROOT}" "$CLAUDE_PROJECT_DIR"

• SessionStart

  bash ${CLAUDE_PLUGIN_ROOT}/scripts/learning-readback.sh

• SessionStart

  bash ${CLAUDE_PLUGIN_ROOT}/scripts/check-plugin-versions.sh

• SessionStart

  bash ${CLAUDE_PLUGIN_ROOT}/scripts/show-notices.sh

Similar plugins