Works with the AI Agents you already use
Spec-Driven Delivery Kit
Run a coding agent like a careful team: written spec, one reviewed batch per branch, independent reviewer subagents, a stop hook that blocks secrets, and reports with measured numbers only.
$12
Spec-Driven Delivery Kit
Example session with this skill installed
Set up this repository for spec-driven delivery, then plan batch 1 from specs/spec.md.
- Read your context and instructions
- Compiled the spec-driven delivery kit
Installer preview, then the kit installed (7 files + settings), CLAUDE.md and domain-guard rules drafted with you, the hook verified on a clean tree; then a batch-1 plan (changes, files, tests, out of scope, open questions) waiting for your approval before any code.
Connects securely to your tools. The creator never sees your data.
About this skill
Coding agents are fast, and they grade their own homework. This kit installs a way of working that has been used on a real project with regulated data:
- Spec first, plan approved, then code. Decisions live in the spec, not in the chat. One batch, one branch; the human merges.
- Independent reviews. A
cold-diff-reviewersubagent reads the batch diff with no knowledge of the conversation and ranks findings blocking / to fix / suggestion with file:line. Adomain-guardtemplate checks your non-negotiable rules in the code, and aneval-runnertemplate compares your evaluation with its baseline. Findings are triaged, two review rounds maximum. - A stop hook that refuses to end the turn while an API key, private key, database URL with credentials or one of your forbidden patterns sits in the files (values never printed, configurable, with a loop guard).
- No unmeasured numbers. A
no-fake-metricsskill: every figure comes with the command that produced it; variable results as a median of 3 runs. - Short end-of-batch reports and a handoff note, so a new session restarts from 20 lines instead of the whole history.
- Safe installer: preview by default, never overwrites an existing file, prints the hook block to merge if you already have a settings file.
Standard-library Python only. The secret patterns in the hook are detection rules, not secrets.
The stop hook is regex-based: it misses secrets in unknown formats and can flag documented examples (use allow_files). The domain-guard and eval-runner agents are templates you fill for your project. Subagents and hooks rely on Claude Code; with other agents, the workflow and templates still apply but the hook and subagents must be wired by hand. The hook command uses python; change it to python3 where needed.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
Requires
Permissions
File Scopes
No network access. The hook prints file names and the type of problem, never the values found.