imbue
Workflow methodologies for analysis, evidence gathering, and structured output.
Overview
Imbue provides reusable patterns for approaching analysis tasks. It’s a methodology plugin - the patterns apply to various inputs (git diffs, specs, logs) and chain together for complex workflows.
Core Philosophy: “NO IMPLEMENTATION WITHOUT A FAILING TEST FIRST” - The Iron Law enforced through proof-of-work validation.
Installation
/plugin install imbue@claude-night-market
Principles
- Generalizable: Patterns work across different input types
- Composable: Skills chain together naturally
- Evidence-based: Emphasizes capturing proof for reproducibility
- TDD-First: Iron Law enforcement prevents cargo cult testing
Skills
Review Patterns
| Skill | Description | When to Use |
|---|---|---|
review-core | Scaffolding for detailed reviews | Starting architecture, security, or code quality reviews |
structured-output | Output formatting patterns | Preparing final reports |
Analysis Methods
| Skill | Description | When to Use |
|---|---|---|
diff-analysis | Semantic changeset analysis | Understanding impact of changes |
catchup | Context recovery | Getting up to speed after time away |
Workflow Guards
| Skill | Description | When to Use |
|---|---|---|
scope-guard | Anti-overengineering with RICE+WSJF scoring | Evaluating features, sprint planning, roadmap reviews |
proof-of-work | Evidence-based validation with output-contracts, retry-protocol, and independent-verification modules | Enforcing Iron Law TDD discipline |
rigorous-reasoning | Anti-sycophancy guardrails | Analyzing conflicts, evaluating contested claims |
assisted-mastery | Assistance-dilemma resolution: visible reasoning, tradeoff surfacing, and fading help | Reviewing or learning from agent-written code |
graduated-implementation | Bounded start, ramp ambition a notch only on demonstrated competence of the prior increment | Building a feature the human must understand, not just ship |
dependency-verification | Package-hallucination and slopsquat defense | Adding, recommending, or installing any package |
Workflow Automation
| Skill | Description | When to Use |
|---|---|---|
workflow-monitor | Execution monitoring and issue creation | After workflow failures or inefficiencies |
Commands
| Command | Description |
|---|---|
/catchup | Quick context recovery from recent changes |
/structured-review | Start structured review workflow with evidence logging |
Agents
| Agent | Description |
|---|---|
review-analyst | Autonomous structured reviews with evidence gathering |
Hooks
| Hook | Type | Description |
|---|---|---|
session-start.sh | SessionStart | Initializes scope-guard, Iron Law, and learning mode |
user-prompt-submit.sh | UserPromptSubmit | Validates prompts against scope thresholds |
tdd_bdd_gate.py | PreToolUse | Enforces Iron Law at write-time |
guard_package_hallucination.py | PreToolUse (Bash) | Flags installs of hallucinated or typosquatted packages; warns by default, blocks with VOW_SHADOW_MODE=0 |
guard_scope_ramp.py | PreToolUse (Write/Edit/MultiEdit) | Holds each increment to the current rung; ramps a notch per recorded demonstration; warns by default, blocks with VOW_SHADOW_MODE=0 |
vow_bounded_reads.py | PreToolUse (Read/Grep/Glob) | Warns or blocks when discovery read budget (15 reads) is exceeded; set VOW_SHADOW_MODE=0 to block |
vow_bounded_reads_reset.py | PreToolUse (Write/Edit/MultiEdit) | Resets the per-session read counter when implementation begins |
vow_no_ai_attribution.py | PreToolUse | Blocks AI attribution strings in git commit messages |
vow_no_edit_narration.py | PreToolUse (Write/Edit/MultiEdit) | Flags text that narrates the edit being made (“We renamed the handler”); reads inserted content only; warns by default, blocks with VOW_SHADOW_MODE=0 |
vow_no_emoji_commits.py | PreToolUse | Blocks emoji characters in git commit messages |
pre-pr-scope-check.sh | Manual | Checks scope before PR creation |
proof-enforcement.md | Design | Iron Law TDD compliance enforcement |
Usage Examples
Structured Review
Skill(imbue:review-core)
# Required TodoWrite items:
# 1. review-core:context-established
# 2. review-core:scope-inventoried
# 3. review-core:evidence-captured
# 4. review-core:deliverables-structured
# 5. review-core:contingencies-documented
Diff Analysis
Skill(imbue:diff-analysis)
# Answers: "What changed and why does it matter?"
# - Categorizes changes by function
# - Assesses risks
# - Summarizes implications
Quick Catchup
/catchup
# Summarizes:
# - Recent commits
# - Changed files
# - Key decisions
# - Action items
Scope Guard
The scope-guard skill prevents overengineering via four components:
| Component | Purpose |
|---|---|
decision-framework | Worthiness formula and scoring |
anti-overengineering | Rules to prevent scope creep |
branch-management | Threshold monitoring (lines, commits, days) |
github-integration | Issue creation and optional Discussion linking for deferrals |
baseline-scenarios | Validated test scenarios |
Iron Law TDD Enforcement
The proof-of-work skill enforces the Iron Law:
NO IMPLEMENTATION WITHOUT A FAILING TEST FIRST
This prevents “Cargo Cult TDD” where tests validate pre-conceived implementations.
RED-Phase Constraints
- The failing test comes before the implementation plan, because a plan written first decides the design the test was meant to find.
- Document the failure before designing tests so that the tests follow the need rather than a pre-conceived implementation.
- A design that looks straightforward still gets its test first: the test is where hidden uncertainty shows up.
TodoWrite Items
proof:iron-law-red - Failing test documented
proof:iron-law-green - Minimal code to pass
proof:iron-law-refactor - Code improved, tests green
proof:iron-law-coverage - Coverage gates verified
See iron-law-enforcement.md module for full enforcement patterns.
Rigorous Reasoning
The rigorous-reasoning skill prevents sycophantic patterns through structured analysis:
| Component | Purpose |
|---|---|
priority-signals | Override principles (no courtesy agreement, checklist over intuition) |
conflict-analysis | Harm/rights checklist for interpersonal conflicts |
debate-methodology | Truth claims and contested territory handling |
red-flag monitoring | Detect sycophantic thought patterns |
Red Flag Self-Check
- “I agree that…” follows the harm/rights checklist, because agreement without validation reports only confidence.
- “You’re right that…” follows evidence that it is proven.
- “That’s a fair point” names the standard it is fair by.
TodoWrite Integration
All skills output TodoWrite items for progress tracking:
review-core:context-established
review-core:scope-inventoried
diff-analysis:baseline-established
diff-analysis:changes-categorized
catchup:context-confirmed
catchup:delta-captured
Integration Pattern
Imbue is foundational - other plugins build on it:
# Sanctum uses imbue for review patterns
Skill(imbue:review-core)
Skill(sanctum:git-workspace-review)
# Pensive uses imbue for evidence gathering
Skill(imbue:proof-of-work)
Skill(pensive:architecture-review)
Superpowers Integration
| Skill | Enhancement |
|---|---|
scope-guard | Uses brainstorming, writing-plans, execute-plan |
Related Plugins
- sanctum: Uses imbue for review scaffolding
- pensive: Uses imbue for evidence gathering
- spec-kit: Uses imbue for analysis patterns