What is deliver-task?

An end-to-end workflow for taking one ready GitHub issue from implementation to a reviewed pull request.

Use it with a ready issue number to assign the work, create the change in a separate worktree, request independent review, and prepare the pull request.

It coordinates coding, an independent review, and a limited round of fixes so the task is ready for a person to merge.

deliver-task — Skill for Claude Code by blakemartz/delivery-loop

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.

Any agent

npx skills add blakemartz/delivery-loop --skill deliver-task

Clone the repo

git clone --depth 1 https://github.com/blakemartz/delivery-loop

Made for: Claude Code.

Or install delivery-loop, the plugin that ships this one along with the rest of its 10 skills.

Wrote 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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/blakemartz/delivery-loop/deliver-task/github.svg)](https://agentmods.dev/skills/blakemartz/delivery-loop/deliver-task)

Your own site

<a href="https://agentmods.dev/skills/blakemartz/delivery-loop/deliver-task"><img src="https://agentmods.dev/badge/skills/blakemartz/delivery-loop/deliver-task/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/blakemartz/delivery-loop/deliver-task"><img src="https://agentmods.dev/badge/skills/blakemartz/delivery-loop/deliver-task.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 66 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 861 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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 5.1	$0.00066	$0.00861
Opus 5.5	$0.00026	$0.00344
Sonnet 5.5	$0.00013	$0.00172
Haiku 4.5	$0.00007	$0.00086

Measured 1mo ago against content hash 4699c7fe1440, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

deliver-task 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.

plugins/delivery-loop/skills/deliver-task/SKILL.md · 38 lines

How it starts

The opening of the file, as written. The whole thing — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.

deliver-task — one issue → a reviewed PR

You drive a single task through the whole loop yourself by dispatching the other roles as subagents (the Agent tool). One task, one branch, one PR, reviewed by a fresh agent, merged by a human. This is the single-task, Level-0 path (runbook §3); for continuous operation use delivery-tick.

Read .claude/delivery.conf for NOTIFY_ESCALATE_CMD. The sub-skills read the rest of the config themselves.

Input

The issue number to deliver (e.g. /delivery-loop:deliver-task 42). It must be a READY task. No argument → report that you need an issue number and stop.

Independence (the reason this is a composition, not one long session)

The reviewer must be a fresh agent with no implementer context. You spawn it as a separate Agent call whose prompt carries only the PR number — never your implementation reasoning. Never review the change yourself. That structural independence is the whole point (spec §3.4).

Protocol

Implement. Spawn an agent: "Run the /delivery-loop:next-task skill on issue #. Report the PR number if you opened one, or why you stopped (claim lost, needs-human filed, queue conflict)."
- If it did not open a PR (claim lost, blocked → needs-human, queue conflict), report that outcome and STOP — there is nothing to review.
Review → patch loop, at most 3 iterations. For round i = 1..3:
1. Spawn a fresh agent: "Run the /delivery-loop:review-task skill on PR #. You have NO prior context on this change — that independence is required and deliberate. Execute every acceptance criterion in a clean checkout, then approve or request changes via gh. Report the URL of the review you actually posted; if you could not post it, say so explicitly."
2. Unposted verdict = escalate. If the reviewer reached a verdict but could not post it to the PR, do not trust it (an unposted review did not happen — spec §11, runbook §7). Label the PR and issue needs-human, comment that the review could not be posted, run NOTIFY_ESCALATE_CMD if it is set, and STOP.
3. Approved → report "PR # ready for human merge" (with round number and the review URL) and STOP. You never merge.
4. Unreviewable (e.g. CI red) → treat like changes-requested; proceed to patch.
5. If this was round 3, break to step 3 (escalation).
6. Patch. Spawn an agent: "Run the /delivery-loop:patch-task skill on PR #." If it escalates (or dies), label needs-human, run NOTIFY_ESCALATE_CMD if set, report, and STOP. Otherwise continue to round i+1.

Read the full file on GitHub · 38 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.

1mo ago First seen · 38 lines · 66 tokens per session scan A 4699c7fe1440

Subscribe to this mod's changes

deliver-task is a skill published in the GitHub repository blakemartz/delivery-loop (5 stars, last pushed 3mo ago), licensed MIT. It adds 66 tokens to every session and 861 once invoked, about $0.0003 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 2026-08-31.

Claude Code Code review Project management agentic ai-agents claude-code-plugin developer-tools

pm-tech-spec-review

Review an engineering technical spec through a PM lens — does it cover every PRD requirement, does it preserve user/UX intent, does it quietly add or drop scope, what are the product risks, what's missing for launch. Produces a priority-tagged findings list + clarifying questions for engineering. Use after eng drafts…

VandanaAjayDubey111/great-pm · 83 tokens

story-analysis

Use when the user provides a user story, ticket, or requirement to analyze before building — acts as a Product Owner. INVEST check, ambiguity hunt, acceptance criteria extraction, epic splitting. Interactive; asks the human one question at a time and never assumes missing details.

jdanigo/hydraia · 56 tokens

deliver-task