A command that implements an approved engineering plan in code, runs the project's checks, and reports the resulting changes.

Use it after a plan has been approved to write the implementation, run toolchain and project gates, and review the resulting diff. It supports new interface work when the relevant design guidance applies.

It keeps coding tied to an approved plan and verifies the work before handing back a summary.

dev — Command by Pierry

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/pierry/harness-kit/dev

Clone the repo

git clone --depth 1 https://github.com/Pierry/harness-kit

Made for: Claude Code.

Per session 17 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 1,481 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found 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.

Model	Per session	Once invoked
Fable 5	$0.00017	$0.01481
Opus 5	$0.00009	$0.00740
Sonnet 5	$0.00003	$0.00296
Haiku 4.5	$0.00002	$0.00148

Measured yesterday against content hash b19b605fdf68, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

dev 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 yesterday.

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.

.claude/commands/sse/dev.md · 101 lines

How it starts

The opening of the file, as written. The whole thing — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Implement plan. Follow .claude/agents/staff-software-engineer/guides/pipeline.md and .claude/shared/pipeline-pattern.md for inputs (resolve-mark-proceed) and eval (adversarial).

Print header card before coding and footer card after gates run. Format: .claude/scripts/stage-card.md.

Source plan: latest in .claude/runtime/outputs/sse/plan/ with approved marker. None, abort. Ask user to run /sse:plan first.

Before coding, write the phase start marker by running this script. Do NOT inline date/printf:

.claude/scripts/marker.sh start .claude/runtime/outputs/sse/.markers/{feature_id}.dev-generate.start

Probe the toolchain with the committed script (do NOT inline node -v/npm ping/$(...) env checks):

.claude/scripts/preflight.sh node npm git

Read:

source plan
source PRP: .claude/runtime/outputs/pm/prp/{feature_id}.md (for the UI signal and product context)
area skill: .claude/agents/staff-software-engineer/skills/{area}/SKILL.md
designer skill: .claude/agents/staff-software-engineer/skills/designer/SKILL.md (when building a new UI: new app, page, feature, or landing). Apply M3 tokens, dark/light theme, modern font, i18n en/pt-BR/es, context-aware favicon.
.claude/agents/staff-software-engineer/guides/coding-style.md
.claude/agents/staff-software-engineer/guides/commit-style.md
project conventions: {repo}/.claude/conventions/{area}.md if present

Read 3+ similar files in target repo before writing. Match conventions:

framework version (Spring 4.3 no-Boot vs Spring Boot, Vue 2 vs 3)
build tool (Maven, Gradle, npm)
package layout, test framework, helpers

Designer-skill gate (decide before any code, by inference not by asking). Determine whether the plan or PRP describes new or changed UI (app, page, feature, landing, component, screen, any visual surface) by reading them (resolve-mark-proceed). Do NOT call AskUserQuestion.

Clearly UI → read + apply designer SKILL.md fully. Repo already has a design system → match it, not imposing M3. Greenfield UI → apply the designer defaults (Material Design 3, dark/light theme, modern font, modern icons, Behance-grade polish, context-aware favicon, i18n, animations).
Clearly not UI (pure backend/devops/no-UI work) → skip the designer skill.
Genuinely ambiguous after reading both → mark NOT FOUND - NEEDS REVIEW: designer-skill applicability and proceed WITHOUT applying it. Either way, still respect the no-emoji and no-em-dash hard rules.

Write code in small commits (1-4 files, < 100 lines ideal). Conventional Commits format.

Code gates (run after each implementation step). One is machine-run, two are yours to apply:

Run this. It executes the repo's own tooling (lint, typecheck, ruff, ktlint, checkstyle, gitleaks):

.claude/runtime/scripts/staff-software-engineer/run-sensors.sh {any-artifact} .claude/agents/staff-software-engineer/sensors/code-maintainability.md

Exit 4 means the repo configures no tooling it knows. That is not a pass, nothing was checked. Report it as a gap; do not claim the gate passed.

Apply these yourself, they need judgment no script has (Execution: inferential). Read them with the Read tool, never cat in a loop. You are the sensor here, so say plainly which you applied and what you found; claiming a pass you did not verify is the failure these gates exist to prevent:

.claude/agents/staff-software-engineer/sensors/code-conventions.md (conventions, banned patterns)
.claude/agents/staff-software-engineer/sensors/test-coverage.md (every feature/bugfix has tests)

Any gate fails, fix and retry. Max 3 attempts. Hard stop after 3.

After done, write .claude/runtime/outputs/sse/dev/{feature_id}.md with summary:

files changed
commits
gate results
blockers if any

Document gates (run on saved summary):

Sensor: .claude/agents/staff-software-engineer/sensors/dev-structure.md (auto-run by post-write hook)
Eval: .claude/agents/staff-software-engineer/evals/dev-quality.md (scored adversarially; threshold 8.0)

Read the full file on GitHub · 101 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.

yesterday First seen · 101 lines · 17 tokens per session scan A b19b605fdf68

Subscribe to this mod's changes

dev is a command published in the GitHub repository Pierry/harness-kit (3 stars, last pushed 1mo ago), licensed MIT. It adds 17 tokens to every session and 1,481 once invoked, about $0.0001 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.

Code review Planning agentic-workflows ai-agents ai-tools anthropic claude claude-code

dev

Getting it into your agent

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

What this file has done since we first saw it

Other commands, from other repositories

setup-pm-skills

statusbar-style

fest-show

superpowers-execute

config

deps-age