harness-kit
01Plugin Claude Code
Harness Kit — a project-agnostic harness-engineering plugin for Claude Code.
5 1mo ago A
tokens not measured
original MIT
harness-kit
02Plugin Claude Code
Project-agnostic harness engineering for Claude Code: session handoff, plan-gating, pre-completion verification gates, loop detection, and maturity-phase routing — all driven by a per-project .harness/config.json. Init into godot or web presets.
5 1mo ago A
tokens not measured
original MIT
evaluator
03Agent
Skeptical, fresh-context evaluator for Generator/Evaluator separation. Scores the current change against the project's rubric (.harness/rubric.md) on each dimension 1-5, running the config.verificationRecipe checks. Any dimension <3 = FAIL. Use before declaring a coding task done. It cannot edit files, so it never…
5 1mo ago A 77 tokens
original MIT
advisor
04Command
Show the project's current harness maturity phase, the artifact metrics behind it, and which harness capabilities are unlocked. Use at session start or when deciding whether to adopt heavier harness practices (evaluator, golden regression, etc.).
5 1mo ago A 43 tokens
original MIT
evaluate
05Command
Dispatch the skeptical evaluator subagent to score the current change against the project rubric (Generator/Evaluator separation). Use before declaring a coding task done, or when the Stop gate recommends it.
5 1mo ago A 37 tokens
original MIT
init
06Command
Initialize harness-kit in the current project. Detects or asks the project type (godot | web | custom) and language (en | zh), scaffolds .harness/config.json + rubric + plan/docs skeleton + a ToC-style CLAUDE.md when the project has none, and enables the git pre-commit gate; then analyzes the codebase to fill in the…
5 1mo ago A 96 tokens
original MIT
plan
07Command
Start the Plan→Build→Verify→Done workflow for a non-trivial change. Enters plan mode seeded with the project's plan template; on approval, the ExitPlanMode hook persists the plan into the configured plan directory. Use before any change larger than 20 lines or touching planned code paths.
5 1mo ago A 60 tokens
original MIT
Command
Analyze the harness's own session trace for failure patterns — verify pass/fail rate, session imbalance, and high-churn (doom-loop) files. Use periodically to decide how to tune the harness itself (thresholds, gates, reasoning budget).
5 1mo ago A 49 tokens
original MIT
verify
09Command
Run the project's harness verification gates (the config-driven harness-verify orchestrator) and report per-gate pass/fail. The manual counterpart to the automatic Stop-hook gate. Use to check work mid-task before declaring done.
5 1mo ago A 45 tokens
original MIT
SessionStart
10Hook
Runs when a session starts, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
original MIT
PreToolUse
11Hook
Runs before the agent uses a tool for Edit, Write and MultiEdit tool calls, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
copy · 56% MIT
PostToolUse
12Hook
Runs after a tool call finishes for ExitPlanMode, Edit, Write and MultiEdit tool calls, running an inline shell check (3 commands). From whieet/harness-kit.
5 1mo ago A
tokens not measured
original MIT
Hook
Runs when a tool call fails, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
original MIT
Hook
Runs when you submit a prompt, before the agent sees it, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
original MIT
PreCompact
15Hook
Runs before the context is compacted, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
original MIT
Setup
16Hook
Runs once when the project is set up for maintenance tool calls, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
original MIT
Stop
17Hook
Runs when the agent finishes a response, running an inline shell check. From whieet/harness-kit.
5 1mo ago A
tokens not measured
copy · 45% MIT
advisor
18Skill Claude CodeCodex
Show the project's current harness maturity phase, the artifact metrics behind it, and which harness capabilities are unlocked. Use at session start or when deciding whether to adopt heavier harness practices (evaluator, golden regression, etc.).
5 1mo ago A 45 tokens
original MIT
evaluate
19Skill Claude CodeCodex
Dispatch the skeptical evaluator subagent to score the current change against the project rubric (Generator/Evaluator separation). Use before declaring a coding task done, or when the Stop gate recommends it.
5 1mo ago A 39 tokens
original MIT
init
20Skill Claude CodeCodex
Initialize harness-kit in the current project. Detects or asks the project type (godot | web | custom) and language (en | zh), scaffolds .harness/config.json + rubric + plan/docs skeleton + a ToC-style CLAUDE.md when the project has none, and enables the git pre-commit gate; then analyzes the codebase to fill in the…
5 1mo ago A 98 tokens
original MIT
plan
21Skill Claude CodeCodex
Start the Plan→Build→Verify→Done workflow for a non-trivial change. Enters plan mode seeded with the project's plan template; on approval, the ExitPlanMode hook persists the plan into the configured plan directory. Use before any change larger than 20 lines or touching planned code paths.
5 1mo ago A 62 tokens
original MIT
Skill Claude CodeCodex
Analyze the harness's own session trace for failure patterns — verify pass/fail rate, session imbalance, and high-churn (doom-loop) files. Use periodically to decide how to tune the harness itself (thresholds, gates, reasoning budget).
5 1mo ago A 53 tokens
original MIT
verify
23Skill Claude CodeCodex
Run the project's harness verification gates (the config-driven harness-verify orchestrator) and report per-gate pass/fail. The manual counterpart to the automatic Stop-hook gate. Use to check work mid-task before declaring done.
5 1mo ago A 47 tokens
original MIT