spec-impl

A command for carrying out a software plan after its decisions have already been made. It uses multiple AI models for reference, but requires their suggestions to be rewritten and reviewed before use.

In plain words
What is it for?
Use it to implement the tasks in an OpenSpec proposal, request code suggestions as patches, and check for side effects before changing files.
Why use it?
It reduces accidental scope changes and prevents unreviewed AI-generated code from being applied directly. It also keeps task files in a format the OpenSpec tool can read.

Command

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 commands/qinye6/pi-ccg/spec-impl
Clone the repo
git clone --depth 1 https://github.com/qinye6/pi-ccg
Per session 12 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,891 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin 100% copy Near-identical to another mod 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 $0.00012 $0.01891
Opus 5 $0.00006 $0.00945
Sonnet 5 $0.00002 $0.00378
Haiku 4.5 $0.00001 $0.00189

Measured 2d ago against content hash d0a5abfd725c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

spec-impl 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.

Origin

This is a copy

100% identical to spec-impl — 6 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.

templates/commands/spec-impl.md · 144 lines

How it starts

The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Core Philosophy

  • Implementation is pure mechanical execution—all decisions were made in Plan phase.
  • External model outputs are prototypes only; must be rewritten to production-grade code.
  • Keep changes tightly scoped; enforce side-effect review before any modification.
  • Minimize documentation—prefer self-explanatory code over comments.

Guardrails

  • NEVER apply 后端/前端模型 prototypes directly—all outputs are reference only.
  • MANDATORY: Request unified diff patch format from external models; they have zero write permission.
  • Keep implementation strictly within tasks.md scope—no scope creep.
  • Refer to openspec/config.yaml for conventions.
  • USER GUIDANCE RULE: When suggesting next steps to the user, ALWAYS use CCG commands (/ccg:spec-research, /ccg:spec-plan, /ccg:spec-impl, /ccg:spec-review). NEVER suggest /opsx:* commands to the user. If OpenSpec CLI returns error messages referencing OPSX skills, translate them to CCG equivalents.
  • TASKS FORMAT RULE: When generating or modifying tasks.md, ALL tasks MUST use checkbox format (- [ ] X.Y description). Heading+bullet format will cause OpenSpec CLI to parse 0 tasks and block the workflow.

Steps

  1. Select Change

    • Run openspec list --json to inspect Active Changes.
    • Confirm with user which change ID to implement.
    • Run openspec status --change "<change_id>" --json to review tasks.
  2. Apply OPSX Change (Pre-flight Check)

    • Call /opsx:apply internally to enter implementation mode:
      /opsx:apply
      
    • This will load the change context and guide you through the tasks defined in tasks.md.
    • Note: This is an internal call. If this step fails, guide the user to re-run /ccg:spec-impl.
    • HARD GATE: Check the returned state field:
      • If state: "blocked" → STOP immediately. Inform the user which artifacts are missing and suggest: "Run /ccg:spec-plan to generate missing artifacts first."
      • If progress.total === 0 → STOP immediately. Inform: "tasks.md has no parseable tasks. Run /ccg:spec-plan to regenerate."
      • Only proceed to Step 3 when state: "ready" and progress.total > 0.
  3. Identify Minimal Verifiable Phase

    • Review tasks.md and identify the smallest verifiable phase.
    • Do NOT complete all tasks at once—control context window.
    • Announce: "Implementing Phase X: [task group name]"
  4. Route Tasks to Appropriate Model

    • Route A: {{FRONTEND_PRIMARY}} — Frontend/UI/styling (CSS, React, Vue, HTML, components)
    • Route B: {{BACKEND_PRIMARY}} — Backend/logic/algorithm (API, data processing, business logic)

    工作目录{{WORKDIR}} 必须通过 Bash 执行 pwd(Unix)或 cd(Windows CMD)获取当前工作目录的绝对路径,禁止从 $HOME 或环境变量推断。如果用户通过 /add-dir 添加了多个工作区,先确定任务相关的工作区。

    For each task:

    codeagent-wrapper --progress --backend <{{BACKEND_PRIMARY}}|{{FRONTEND_PRIMARY}}> {{GEMINI_MODEL_FLAG}}{{GROK_MODEL_FLAG}}- "{{WORKDIR}}" <<'EOF'
    TASK: <task description from tasks.md>
    CONTEXT: <relevant code context>
    CONSTRAINTS: <constraints from spec>
    OUTPUT: Unified Diff Patch format ONLY
    EOF
    

    会话复用:保存返回的 SESSION_ID:({{BACKEND_PRIMARY}} → CODEX_PROTO_SESSION,{{FRONTEND_PRIMARY}} → GEMINI_PROTO_SESSION),Step 7 审查时复用。

  5. Rewrite Prototype to Production Code Upon receiving diff patch, NEVER apply directly. Rewrite by:

    • Removing redundancy
    • Ensuring clear naming and simple structure
    • Aligning with project style
    • Eliminating unnecessary comments
    • Verifying no new dependencies introduced
  6. Side-Effect Review (Mandatory before apply) Verify the change:

    • Does not exceed tasks.md scope
    • Does not affect unrelated modules
    • Does not introduce new dependencies
    • Does not break existing interfaces

    If issues found, make targeted corrections.

  7. Multi-Model Review (PARALLEL)

    • CRITICAL: You MUST launch BOTH {{BACKEND_PRIMARY}} AND {{FRONTEND_PRIMARY}} in a SINGLE message with TWO Bash tool calls.
    • DO NOT call one model first and wait. Launch BOTH simultaneously with run_in_background: true.

Read the full file on GitHub · 144 lines

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 · 144 lines · 12 tokens per session scan A d0a5abfd725c

Subscribe to this mod's changes

spec-impl is a command published in the GitHub repository qinye6/pi-ccg (10 stars, last pushed 8d ago), licensed MIT. It adds 12 tokens to every session and 1,891 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to spec-impl, differing in 6 lines, and is treated as a copy.