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.
npx agentmods add skills/w00fx/spec-anchored-agentic-development/implement-featurenpx skills add w00fx/spec-anchored-agentic-development --skill implement-featuregit clone --depth 1 https://github.com/w00fx/spec-anchored-agentic-developmentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/w00fx/spec-anchored-agentic-development/implement-feature)<a href="https://agentmods.dev/skills/w00fx/spec-anchored-agentic-development/implement-feature"><img src="https://agentmods.dev/badge/skills/w00fx/spec-anchored-agentic-development/implement-feature.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00065 | $0.00863 |
| Opus 5 | $0.00032 | $0.00432 |
| Sonnet 5 | $0.00013 | $0.00173 |
| Haiku 4.5 | $0.00006 | $0.00086 |
Grade A, and why
implement-feature 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 3d 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.
How it starts
The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
implement-feature — supervised-local adapter
Phase 0 first action: read .claude/protocols/implementation-protocol.md
NOW. (Installed at user level, the same file sits under your ~/.claude;
the Codex port reads it from agent-system/protocols/. If none of the three
resolves, that is a NAMED_BLOCKER — never run the machine from memory.) It is the state machine — phases, terminals, invariants,
logging, rationalizations. This file only configures who satisfies
each gate. Treat the protocol as mandatory, not as an optional
reference.
Mode identity
The local, human-driven adapter, invoked directly as /implement-feature (or by name); /implement is a redirect that only points here. Runs
interactively — plain. (A /goal wrapper is optional and only as a
separately verified composition: the skill loads by your explicit
invocation, never by being named inside a condition.) It must NEVER run under headless /goal
(claude -p) — no one answers there. Headless work is
implement-backlog; orchestrated workers are implement-orchestrated
(one ticket, one PR — never reroute them here or to backlog).
Scenarios
- New capability from its ratified spec (spec creation itself is human-led — out of scope here).
- Feature within an existing capability.
- Increment: bugfix, tech-debt, rule adjustment, reference-data update proposal, feature-within-scope backlog item.
Gate providers (this mode)
| Gate | Provider |
|---|---|
| Ambiguity (Phase 2) | the human, in-session (AskUserQuestion) |
| Plan approval (Phase 3) | the human, in-session (ExitPlanMode); fingerprint noted in the log |
| Spec change (Phase 4) | semantic-amendment gate in-session (affected IDs, old → proposed meaning, rationale, verification change); on explicit human approval the skill MAY materialize the amendment on this branch — the PR review ratifies it (requires_human_approval). Orchestrated and unattended adapters stay proposal-only |
| Scope expansion | the human, in-session |
| Review (Phase 8) | dispatch the reviewer agent in isolated context (it did not write this code; the agent is model-pinned to Opus at max effort — the judge never runs on a cheaper brain). Confirm a non-empty diff or a NO_CHANGE_CANDIDATE evidence target before dispatch (the no-op claim gets corroborated, never self-declared). Default: one reviewer, all applicable lenses; diffs beyond ~400 lines: parallel single-lens dispatches. Findings are report-only; fix root-cause; re-seal |
| Delivery (Phase 9) | PR opened; the human merges. Park = end the session normally |
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago First seen · 65 lines · 65 tokens per session scan A 3f9f996e2c36
implement-feature is a skill published in the GitHub repository w00fx/spec-anchored-agentic-development (5 stars, last pushed 8d ago), licensed MIT. It adds 65 tokens to every session and 863 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
Accept
The boundary ritual — review what just landed, record an approval, journal it, and advance the phase. The ONLY step that writes approvals, and it does NO git.
A5 — Web Audit
Research the world. Real web research to find what the best teams do for every technology and pattern in the build, then implement superior approaches now (not later). The only lens that brings outside knowledge.
Brainstorm
Interactive ideation and decision capture at any pipeline boundary. Opinionated dialogue that surfaces contradictions, gaps, and untested assumptions — not template-filling. Feeds the Furnace.
Specify
Write the master spec — the WHAT. User stories, functional & non-functional requirements, acceptance criteria, and the dependency-ordered micro-spec plan. Strictly WHAT, never HOW.
A1 — Design & Master-Spec Consistency
The drift check. Compare what was built against the design doc and master spec. Sets the Double Gate — A3 cannot start until A1 is at zero findings.
A3 — Polish
Code/content quality. Dead code, hardcoded values, leftover TODOs, weak error handling, security holes, performance traps, standards violations. Enforces the Double Gate — blocks until A1 is at zero findings.