build

Implements SPEC.md tasks through a plan-then-execute loop with failure backpropagation.

1|Updated May 6, 2026
One-click install
npx skills add https://github.com/surfingalien/FinSurfing --skill build-surfingalien
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: build
Source: https://github.com/surfingalien/FinSurfing/tree/main/.claude/skills/ck%3Abuild
Command: npx skills add https://github.com/surfingalien/FinSurfing --skill build-surfingalien

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a written specification into working code often drifts: tasks get skipped, invariants get violated, and test failures get patched blindly without fixing the root cause. This Skill enforces a disciplined single-threaded workflow that implements a SPEC.md file task by task, tracks status directly in the spec, and routes every failure through a root-cause analysis before retrying. ## Core Features & Use Cases - Spec-driven task execution: Parses §T task rows from SPEC.md and implements them individually (build §T.3), the next pending one (build --next), or all pending tasks in order. - Plan gating: Before writing code, it cites every applicable §V invariant and §I interface, lists files and tests to touch, and waits for user approval. - Failure backpropagation: On test or build failure, it classifies the cause as a code bug, a spec error, or an unspecified edge case, and updates the spec's §V invariants and §B memory before resuming. - Use Case: A developer maintains a living SPEC.md for a project and asks the assistant to "build --next"; the skill flips the task status, implements it, runs the verification command, commits with invariant citations, and moves on. ## Quick Start Ask the assistant to run the build skill with "build --next" to implement the next pending task from SPEC.md.

Frequently Asked Questions about build

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

FAQPage Schema
How do I implement tasks from a SPEC.md file with an AI assistant?

Invoke the build skill with a task selector such as "build §T.3" for one task, "build --next" for the lowest pending task, or no argument to run all pending tasks in order. The skill plans against the spec's invariants, executes, verifies, and updates task status in SPEC.md.

What happens when a test fails during spec-driven implementation?

The skill classifies the failure as a code bug, a spec error, or an unspecified edge case. Code bugs are fixed directly, while spec problems trigger a spec update that adds a new invariant to prevent recurrence before the build resumes.

Does the build skill work without a SPEC.md file?

No. The skill expects SPEC.md to exist and stops immediately if it is missing, directing you to invoke the spec skill first to author the specification before any implementation begins.

Can the build skill edit SPEC.md beyond task status?

No. Its write policy only permits flipping §T task status cells between pending, in-progress, and done. Any other specification change must go through the separate spec skill to keep the spec authoritative.

Does the build skill use sub-agents or parallel workers?

No. It is explicitly single-threaded and runs on the main conversation thread only. It avoids sub-agents, parallel workers, and progress dashboards, using the SPEC.md task table itself as the status view.