core

Executes TDD-driven code changes through gated Absorb, Clarify, Plan, Execute state transitions.

10|2|Updated Jan 24, 2026
One-click install
npx skills add https://github.com/nrdxp/predicate --skill core-nrdxp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: core
Source: https://github.com/nrdxp/predicate/tree/main/skills/core
Command: npx skills add https://github.com/nrdxp/predicate --skill core-nrdxp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long coding sessions drift: agents skip tests, commit without verification, and improvise past ambiguity. This Skill enforces a disciplined micro-execution protocol where every code change is preceded by a failing test, verified by a deterministic evaluator, and gated at each commit boundary. ## Core Features & Use Cases - State-Machine Execution: Moves work through Absorb, Clarify, Plan, and Execute states, halting on ambiguity instead of guessing. - TDD Verification Loop: Requires a baseline failing test before implementation, then iterates corrections (capped at 3-5 loops) until the validator passes. - Commit Gates: Blocks commits until conventional-commit message hygiene and ledger validation pass, with MANUAL (human approval) or AUTOMATIC control modes. - Use Case: Ask the agent to fix a bug with the core workflow; it first writes a failing test reproducing the bug, implements the fix, refines the diff through a skeptical audit loop, and only then proposes a gated commit. ## Quick Start Invoke the core workflow to implement my planned change using test-driven steps with verification and commit gates at each boundary.

Frequently Asked Questions about core

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a test-driven coding workflow with an AI agent?

Invoke the core workflow with your goal and constraints. The agent writes a failing test first to establish a baseline, implements the fix, re-runs the validator until it passes, and halts at each commit boundary for approval.

What happens when the agent encounters ambiguous requirements?

The workflow transitions to the CLARIFY state, populates an OBSTACLES list, and halts all code generation. Implementation is forbidden until the uncertainties are explicitly resolved, following the halt-over-assumption rule.

Can the core workflow commit changes automatically?

Yes, when CONTROL_MODE is set to AUTOMATIC and permissions are active, it commits after tests pass and the ledger gate authorizes the change. The default MANUAL mode outputs a review block and waits for human confirmation.

What happens if tests keep failing during execution?

The verification loop applies corrective edits for up to 3-5 iterations to minimize the error. If no convergence occurs, the workflow transitions back to CLARIFY and halts rather than committing broken code.

When should I not use the core workflow?

Do not use it for high-level design exploration or architecture changes. The protocol explicitly exits and halts if the work drifts into exploring design alternatives, which belongs to planning dispositions outside this workflow.