tapd-story-validate

Validates implemented code through four parallel review dimensions and orchestrates in-place fixes.

840|261|Updated Apr 18, 2019
One-click install
npx skills add https://github.com/TencentBlueKing/bk-bcs --skill tapd-story-validate
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tapd-story-validate
Source: https://github.com/TencentBlueKing/bk-bcs/tree/main/.agents/skills/tapd-story-pipeline/tapd-story-validate
Command: npx skills add https://github.com/TencentBlueKing/bk-bcs --skill tapd-story-validate

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

After a requirement is implemented, teams need a structured way to verify the code against architecture, security, code review, and test coverage standards before accepting it. This Skill automates that validation stage of the TAPD story pipeline, running four parallel review subagents and coordinating fixes or rollbacks based on their verdicts.

Core Features & Use Cases

  • Four-Dimension Parallel Validation: Runs architecture, security, CodeReview, and test-coverage checks simultaneously via independent subagents, each producing its own report and log file.
  • In-Place Fix Loop: When findings are attributed to the code itself, dispatches backend-developer or frontend-developer agents based on affected file paths, then re-runs validation (up to 3 rounds per attempt).
  • Code-Preserving Rollback: When specs or plans are insufficient, preserves all code and reports, writes an iteration patch, and re-enters upstream phases with unresolved findings.
  • Use Case: A fullstack feature is implemented on an iteration branch; this Skill validates it in parallel, routes a backend contract mismatch to the backend-developer for fixing, and marks the story phase as validated once all four reports return LGTM.

Quick Start

Run the validate stage for the current implemented story to execute the four parallel validation checks and apply any needed fixes.

Frequently Asked Questions about tapd-story-validate

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

FAQPage Schema
How do I validate implemented code across multiple review dimensions?

Run the validate stage, which launches four parallel subagents for architecture, security, CodeReview, and test coverage. Each produces a report with a verdict; all four must return LGTM with no CRITICAL or HIGH findings for the phase to advance to validated.

How are validation fixes assigned between backend and frontend developers?

Fix routing depends on affected file paths: backend files like .go or .proto go to backend-developer, frontend files like .vue or .ts go to frontend-developer, and mixed changes run serially with backend first. Contract mismatches are attributed by comparing code against plan.md.

What happens when validation finds the spec or plan is insufficient?

The Skill triggers a code-preserving rollback: all review reports and code are kept, an iteration patch is written with unresolved findings, attempts is incremented in meta.yaml, and the pipeline re-enters the upstream phase for spec revision.

How many fix-and-revalidate rounds are allowed per attempt?

A maximum of 3 rounds per attempt is allowed. If CRITICAL or HIGH findings remain after 3 rounds, the stage force-escalates to a rollback re-entry rather than continuing to retry.

Does the validation stage roll back code on failure?

No. The Skill explicitly forbids git reset, stash, checkout, or clean operations. Both round retries and rollback re-entries preserve all code, reports, and logs, with downstream stages working in differential fix mode.