goal-test

A local experiment for testing a goal command that keeps an AI coding session working until a stated condition is judged complete. It uses a separate language model to evaluate the conversation after each assistant turn.

In plain words
What is it for?
Experimenting with long-running, convergent coding tasks and evaluating whether the transcript shows that a defined goal has been met.
Why use it?
It helps test whether an autonomous coding workflow can continue toward a verifiable end state instead of stopping after one incomplete step.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/restarter/lets-workflow/goal-test
Any agent
npx skills add restarter/lets-workflow --skill goal-test
Clone the repo
git clone --depth 1 https://github.com/restarter/lets-workflow

Made for: Claude Code, Codex.

Per session 137 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 5,782 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00137 $0.05782
Opus 5 $0.00068 $0.02891
Sonnet 5 $0.00027 $0.01156
Haiku 4.5 $0.00014 $0.00578

Measured 2d ago against content hash cda95a72bb68, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

goal-test scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/goal-test/SKILL.md · 281 lines

How it starts

The opening of the file, as written. The whole thing — 281 lines — stays where its author put it; the contents beside it link to each section on GitHub.

goal-test (local experiment)

/goal installs a session-scoped Stop hook with an LLM-evaluated termination condition. After every model turn, an evaluator-LLM reads the transcript and judges whether the goal is met. If not, the turn is rejected and the model must keep working. Best fit: convergent autonomous work (drive a task / epic / PR to a defined end state).

Not for periodic polling — that's /loop's job (see loop-test skill).

How /goal actually works

Sources: claude binary v2.1.150 strings + official docs (code.claude.com/docs/en/goal) + community write-ups (see References).

Mechanics

  1. User runs /goal <condition>. Stored as session-scoped state. Cap: 4000 characters per condition.
  2. After every model turn (assistant message that would normally yield), evaluator-LLM is invoked with the condition + transcript.
    • Evaluator runs on your configured "small fast model" (Haiku by default). Eval tokens are billed separately; Anthropic describes them as "typically negligible" but they accumulate on long runs.
    • Evaluator reads the transcript only — assistant text + user messages. Tool inputs/outputs ARE in transcript, but you must surface evidence in your assistant text for the evaluator to weight it heavily. "Quote specific text from transcript whenever possible" is in its system prompt.
  3. Evaluator returns JSON:
    • {"ok": true, "reason": "<quote evidence>"} → goal achieved, session yields, UI shows ✔ Goal achieved (Ns · N turn · N tokens) (user-visible only; not surfaced into assistant transcript — empirically confirmed Experiment 1).
    • {"ok": false, "reason": "<why not>"} → Stop hook rejects, model gets a system message and must continue. Common reason text: insufficient evidence in transcript.
    • {"ok": false, "impossible": true, "reason": "..."} → only when genuinely unachievable (self-contradictory condition, missing resource, exhausted approaches). Evaluator is instructed to NOT trust the model's self-assessment — it must independently judge whether the goal is structurally impossible.
  4. Hard cap: CLAUDE_CODE_STOP_HOOK_BLOCK_CAP env var (present in binary, undocumented publicly; the community-cited "500" comes from the unofficial jthack/claude-goal precursor, not from official /goal). After N blocks the loop force-yields. Actual default unknown.

Read the full file on GitHub · 281 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 281 lines · 137 tokens per session scan A cda95a72bb68

Subscribe to this mod's changes

goal-test is a skill published in the GitHub repository restarter/lets-workflow (17 stars, last pushed 9d ago), licensed MIT. It adds 137 tokens to every session and 5,782 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

obra/superpowers · 37 tokens

chat-pet-sprite-creation

Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.

microsoft/vscode · 53 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens