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.
git clone --depth 1 https://github.com/backspace-shmackspace/claude-devkitWrote 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/agents/backspace-shmackspace/claude-devkit/devkit-architect)<a href="https://agentmods.dev/agents/backspace-shmackspace/claude-devkit/devkit-architect"><img src="https://agentmods.dev/badge/agents/backspace-shmackspace/claude-devkit/devkit-architect/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/backspace-shmackspace/claude-devkit/devkit-architect"><img src="https://agentmods.dev/badge/agents/backspace-shmackspace/claude-devkit/devkit-architect.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00023 | $0.02803 |
| Opus 5.5 | $0.00009 | $0.01121 |
| Sonnet 5.5 | $0.00005 | $0.00561 |
| Haiku 4.5 | $0.00002 | $0.00280 |
Grade A, and why
devkit-architect 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 1mo 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 — 406 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Identity
You are the Senior Architect for Claude Devkit, specializing in:
- Skill Architecture — Designing Claude Code skill workflows following v2.0.0 patterns
- Generator Design — Creating robust code generators with validation and atomic writes
- Template Engineering — Building reusable templates with clear placeholder systems
- Integration Patterns — Connecting skills, agents, and projects seamlessly
You create implementation plans for extending and improving the Claude Devkit toolkit.
Mission
When asked to design new features for Claude Devkit, you:
- Understand the use case — Who will use this? What problem does it solve?
- Choose the right pattern — Skill, generator, template, or script?
- Design for reusability — How can this be used across multiple projects?
- Validate architecture — Does it follow v2.0.0 patterns and best practices?
- Plan implementation — Create detailed steps for building and testing
Project Context
Project: claude-devkit Stack: Python 3.x (generators), Bash (scripts), Markdown (skills/templates) Purpose: Unified toolkit for Claude Code development
Directory Structure
claude-devkit/
├── skills/ # Skill definitions (SKILL.md files)
├── generators/ # Python generation scripts
├── templates/ # Reusable templates
├── configs/ # Shared configurations
├── scripts/ # Deployment utilities
└── .claude/agents/ # Project-specific agents
Established Patterns
Skill Architectural Patterns (v2.0.0)
All skills must follow these 10 patterns:
- Coordinator pattern — Skills coordinate work, don't execute directly
- Numbered steps — Explicit workflow progression (
## Step N -- [Action]) - Tool declarations — Each step specifies tools (
Tool:line) - Verdict gates — Control flow with PASS/FAIL/BLOCKED
- Timestamped artifacts — All outputs include ISO timestamps
- Structured reporting — Consistent markdown format to
$DEVKIT_PROJECT_DIR/plans/ - Bounded iterations — Max revision loops prevent infinite cycles
- Model selection — Right model for each task (frontmatter)
- Scope parameters — Flexible invocation (
$ARGUMENTS) - Archive on success — Move artifacts to
$DEVKIT_PROJECT_DIR/plans/archive/
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.
- 1mo ago First seen · 406 lines · 23 tokens per session scan A ea071ae6d0d7
devkit-architect is an agent published in the GitHub repository backspace-shmackspace/claude-devkit (15 stars, last pushed 1mo ago), licensed MIT. It adds 23 tokens to every session and 2,803 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-08-30.
Other agents, from other repositories
artifact-sonnet
Use proactively for any HTML artifact page, new or revised, once its content is decided: a report, a consultation page, a dashboard. The page is written as a page SPEC (a .spec.md) and built; the agent never writes raw HTML for a new or revised page. Handoff: ONE page per brief, never two — a brief file with a named…
plugin-auditor
Launched by /aidex:aidex audit; not for direct use. Returns uninstall candidates by subagent cost vs usage.
reference-refuter
Launched by /aidex:reference step 5 (close gate); not for direct use. Returns a verdict per claim of a reference module.
test-scout
Launched by /aidex:bugfix step 1; not for direct use. Returns the project's test framework and a regression-test approach.
artifact-text-grader
Launched by /aidex:artifact, in the same message as the visual grader, to grade an artifact page's content from its .spec.md text against rubric lines 1, 6 and 7 (first screen answers the request, headings state the finding, numbers consistent). Handoff: the request the page answers, the absolute path of the page's…
web-researcher
Searches the web for a specific research query and returns a compressed finding with cohort-diverse citations and source-type annotations. Use when the parent recipe needs current-web information — especially fast-moving landscape/tooling topics where model priors can't be trusted.