Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add MadAppGang/magus/plugin install devWrote 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/commands/madappgang/magus/release)<a href="https://agentmods.dev/commands/madappgang/magus/release"><img src="https://agentmods.dev/badge/commands/madappgang/magus/release/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/commands/madappgang/magus/release"><img src="https://agentmods.dev/badge/commands/madappgang/magus/release.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 | $0.00050 | $0.06676 |
| Opus 5.5 | $0.00020 | $0.02670 |
| Sonnet 5.5 | $0.00010 | $0.01335 |
| Haiku 4.5 | $0.00005 | $0.00668 |
Grade A, and why
release scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 22d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 442 lines — stays where its author put it; the contents beside it link to each section on GitHub.
<user_request>$ARGUMENTS</user_request>
<golden_rules> These four rules override convenience at every phase. They are the distilled failure history of release-please, changesets, semantic-release, cargo-release and npm itself.
- REVERSIBLE FIRST, IRREVERSIBLE LAST. No tag leaves the machine and no publish starts until every gate has passed on the exact SHA being released.
- THE TESTED CANDIDATE IS THE RELEASED CANDIDATE. Changing versions, changelogs, generated files, or lockfiles after tests ran means the tested candidate is no longer the candidate being released. Any post-test mutation returns the flow to PREPARE.
- ONE HUMAN DECISION, MADE UP FRONT. Authorization is settled BEFORE the pipeline starts (see ), never mid-run. If the user said "release yourself", run to completion without stopping; if they did not, ask once, right away, before doing anything. A release that blocks in the middle waiting for a click is the failure mode this rule exists to prevent. The PR stays as the durable, reviewable record — it is not a second approval step.
- EVERY SIDE EFFECT IS A PREDICATE + ACTION. Before acting, check whether the effect already exists: absent → do it; present and matching → skip and say so; present but DIFFERENT → hard stop, this is a consistency incident, never a retry. </golden_rules>
<worktree_context> Releases are EXPECTED to run from a git worktree, not the main checkout. Detect it (git rev-parse --git-dir vs --git-common-dir differ) and treat it as normal, but know what a worktree changes:
- Tags, branches, and remotes are SHARED with the main checkout and every sibling worktree. A tag created here is instantly visible everywhere; the explicit-ref push rule matters double because sibling worktrees accumulate each other's local tags.
- The stash stack is shared too. Never bare
git stash/git stash popduring a release — another session may push or pop concurrently. A temporary WIP commit is the safe way to set something aside. - Sibling worktrees may hold in-flight work on the same files. Release from the last COMMITTED state of this branch; never chase or wait on another session's uncommitted edits.
- THE WORKTREE IS DISPOSABLE. After the release merges, the worktree is routinely reaped — and everything gitignored inside it (session folders, scratch, caches, local state) is deleted without ceremony. That is the designed lifecycle, which is exactly why the knowledge-promotion step in PREPARE and the residue report in REPORT exist: anything worth keeping must leave the worktree as a committed file BEFORE the merge, or it never will. </worktree_context>
- The user's request already says it ("release yourself", "merge and release without asking", "--auto", or equivalent) → AUTONOMOUS MODE for this whole run. Do not ask again at any point. Report every irreversible step as it happens.
- It does not → ask ONE question now, before starting: "Run this release end to end myself (merge + tag + publish without stopping), or stop before the irreversible steps for your go-ahead?" Their answer sets the mode for the run. Then start.
- Never introduce a new approval question in the middle of the pipeline. The only legitimate mid-run stops are NOT approvals: a preflight gate that failed, or a consistency incident (golden rule 4: something exists but differs). Those stop the run because proceeding would be wrong, not because permission is missing.
- Authorization covers this run only. It does not carry over to the next invocation.
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.
- 22d ago Changed · -1 lines 5eb94d950e71
- 1mo ago Changed · +134 lines 33167f781f19
- 1mo ago First seen · 309 lines · 50 tokens per session scan A 6132d1cf4f5a
release is a command published in the GitHub repository MadAppGang/magus (13 stars, last pushed today), licensed MIT. It adds 50 tokens to every session and 6,676 once invoked, about $0.0002 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 2026-09-04.
Other commands, from other repositories
land
Resolve and squash merge one named, currently authorized pull request.
quality
A command for temporarily clearing a release quality gate or one named check, while recording the reason and expiry time. A quality gate is a required check that normally blocks completion or release.
release-notes
Generate consistent, well-structured release notes from git history. Triggered on release tags following semver patterns (v..) to produce categorized changelog with breaking changes, features, fixes, and contributor attribution.
create-pr
Push the current branch and open a pull request into main with a structured description derived from the branch's commits and the open issues it resolves.
release-manager
Automated release coordination and deployment with ruv-swarm orchestration for seamless version management, testing, and deployment across multiple packages.
ship
Ship is the operational release-prep flow for a connected project repo.