How do I install the Goal Bridge skill?

Run npx skillmds@latest add arbazkhan971/goal-bridge in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Goal Bridge safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Goal Bridge?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Goal Bridge free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Goal Bridge?

arbazkhan971 (@arbazkhan971) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Goal Bridge by arbazkhan971 — Agent Skill

Name: Goal Bridge
Author: arbazkhan971

File contents

Goal-Bridge — Completion Contracts for Goal Modes

Activate When

User invokes /godmode:goal-bridge
User sets a goal in a harness goal mode (/goal or your harness's equivalent)
User asks "how do we know this is done"
An evaluator needs a command that exits 0

Workflow

Step 1: Derive the Contract BEFORE Execution

No work starts until the contract exists. Derive all four fields up front:

metric — ONE shell command where exit 0 = goal met. Never prose judgment.
threshold — numeric or boolean bound the metric must satisfy.
evidence — file path where metric output/proof is written every round.
rollback — exact trigger that reverts the work.

Step 2: Emit the Contract Block

Single source of truth — use EXACTLY this shape everywhere:

Goal-Bridge Contract (mandatory final output):
- metric: <single shell command; contract met iff it exits 0>
- threshold: <numeric or boolean bound the metric must satisfy>
- evidence: <file path where metric output/proof is written every round>
- rollback: <exact trigger that reverts the work, e.g. "metric fails on 2 consecutive rounds">

Step 3: Run the Metric Every Round

Each round: run the metric command and append its output to the evidence path.

Step 4: Rollback on Trigger

git reset --hard <last-good-commit>

Then log the attempt: what failed, which round, evidence excerpt.

Step 5: Hand Off the Contract

Give the contract block to the harness goal mode / evaluator (/goal or your harness's equivalent). The evaluator consumes "command exits 0".

Hard Rules

The contract exists BEFORE work starts.
The metric is a single command, never prose judgment.
The evidence file is written every round.
Never declare done without exit 0.

Stop Conditions

STOP when FIRST of:
  - metric exits 0 AND evidence written -> DONE
  - rollback trigger fires -> revert + report PARTIAL
  - cannot express goal as exit-0 command -> report NEEDS_CONTEXT (do NOT fake a metric)

Integration

Works with any harness goal mode: the /goal command (or your harness's equivalent).
pi goal contracts.
Any harness evaluator that consumes "command exits 0".
Philosophy source: skills/verify/SKILL.md.

---name: goal-bridgedescription: Machine-checkable completion contracts for agent goal modes. goal mode, /goal, goal contract, completion contract, evaluator, metric command, exit-zero contract, evidence path, rollback trigger.---# Goal-Bridge — Completion Contracts for Goal Modes## Activate When- User invokes `/godmode:goal-bridge`- User sets a goal in a harness goal mode (`/goal` or your harness's equivalent)- User asks "how do we know this is done"- An evaluator needs a command that exits 0## Workflow### Step 1: Derive the Contract BEFORE ExecutionNo work starts until the contract exists. Derive all four fields up front:- **metric** — ONE shell command where exit 0 = goal met. Never prose judgment.- **threshold** — numeric or boolean bound the metric must satisfy.- **evidence** — file path where metric output/proof is written every round.- **rollback** — exact trigger that reverts the work.### Step 2: Emit the Contract BlockSingle source of truth — use EXACTLY this shape everywhere:```Goal-Bridge Contract (mandatory final output):- metric: <single shell command; contract met iff it exits 0>- threshold: <numeric or boolean bound the metric must satisfy>- evidence: <file path where metric output/proof is written every round>- rollback: <exact trigger that reverts the work, e.g. "metric fails on 2 consecutive rounds">```### Step 3: Run the Metric Every RoundEach round: run the metric command and append its output to the evidence path.### Step 4: Rollback on Trigger```bashgit reset --hard <last-good-commit>```Then log the attempt: what failed, which round, evidence excerpt.### Step 5: Hand Off the ContractGive the contract block to the harness goal mode / evaluator (`/goal` or yourharness's equivalent). The evaluator consumes "command exits 0".## Hard Rules1. The contract exists BEFORE work starts.2. The metric is a single command, never prose judgment.3. The evidence file is written every round.4. Never declare done without exit 0.## Stop Conditions```STOP when FIRST of: - metric exits 0 AND evidence written -> DONE - rollback trigger fires -> revert + report PARTIAL - cannot express goal as exit-0 command -> report NEEDS_CONTEXT (do NOT fake a metric)```## Integration- Works with any harness goal mode: the `/goal` command (or your harness's equivalent).- pi goal contracts.- Any harness evaluator that consumes "command exits 0".- Philosophy source: skills/verify/SKILL.md.

Frequently asked questions

Goal Bridge