maf-release-watcher

maf-release-watcher is a skill for Claude Code, Codex from joslat/maf-doctor. It costs 124 tokens per session (3,157 once invoked), scanned A, original, MIT.

An automated release-tracking workflow for Microsoft Agent Framework, or MAF, a software toolkit that changes across published versions.

In plain words
What is it for?
It extracts release data, records breaking changes, updates version and compatibility files, creates migration guides, and opens pull requests or issues for changes that need human judgment.
Why use it?
It reduces the manual work of checking each new release and keeps compatibility information and migration guidance current.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit It extracts release data, records breaking changes, updates version and compatibility files, creates migration guides, and opens pull requests or issues for changes that need human judgment.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/joslat/maf-doctor/maf-release-watcher
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 joslat/maf-doctor --skill maf-release-watcher
Clone the repo
git clone --depth 1 https://github.com/joslat/maf-doctor

Made for: Claude Code, Codex.

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.

agentmods badge for maf-release-watcher

README.md
[![agentmods](https://agentmods.dev/badge/skills/joslat/maf-doctor/maf-release-watcher/github.svg)](https://agentmods.dev/skills/joslat/maf-doctor/maf-release-watcher)
Your own site
<a href="https://agentmods.dev/skills/joslat/maf-doctor/maf-release-watcher"><img src="https://agentmods.dev/badge/skills/joslat/maf-doctor/maf-release-watcher/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.

agentmods 80×15 button for maf-release-watcher

Your own site · 80×15
<a href="https://agentmods.dev/skills/joslat/maf-doctor/maf-release-watcher"><img src="https://agentmods.dev/badge/skills/joslat/maf-doctor/maf-release-watcher.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 124 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,157 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. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00124 $0.03157
Opus 5 $0.00062 $0.01579
Sonnet 5 $0.00025 $0.00631
Haiku 4.5 $0.00012 $0.00316

Measured 12d ago against content hash 6757a90ab537, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

maf-release-watcher 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 12d 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.

.github/skills/maf-release-watcher/SKILL.md · 199 lines

How it starts

The opening of the file, as written. The whole thing — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.

maf-release-watcher

Purpose

When Microsoft ships a new MAF version, this pipeline keeps the toolkit's three artefacts current with zero manual baseline-update work:

  • .maf-version — the tracked-current pointer
  • docs/compatibility-matrix.md — dependency-version table (new row per release)
  • guides/maf-X.Y.Z-migration-guide.md — per-version delta with banner pointing back to the chain
  • guides/maf-current-migration-guide.md — auto-regenerated cumulative reference
  • .github/skills/maf-obsolete-api-registry/registry.yaml — append-only breaking-change registry

Two of those (compatibility-matrix.md row contents, per-version guide TODO sections, registry-entry TODO fields) require judgement the deterministic pipeline can't fully automate — release notes have to be interpreted, before/after C# examples written, etc. That work is delegated to GitHub Copilot Coding Agent via a second workflow (maf-ai-fill-todos.yml) that opens an issue with a structured prompt and assigns the bot.


When this skill is "invoked"

It isn't, in the Copilot-Chat sense. The work runs in GitHub Actions VMs, not in a Copilot conversation. This skill is the reference doc explaining what those workflows do, so a maintainer reading the repo can understand the whole loop without piecing it together from two .yml files plus three Python scripts.

Triggers for the pipeline itself:

  • Weekly cron: 0 6 * * 4 (Thursday 06:00 UTC) — defined in .github/workflows/maf-release-watcher.yml
  • Manual dispatch: gh workflow run maf-release-watcher.yml -f maf_version=X.Y.Z

Architecture — the three stages

TRIGGER (cron or gh workflow run)
   │
   ▼
┌────────────────────────────────────────────────────────────────────┐
│  STAGE 1 — Deterministic data extraction                           │
│  .github/workflows/maf-release-watcher.yml                         │
│  Runs on a fresh Ubuntu VM provisioned by GitHub Actions           │
│  NO LLM, NO AGENT — just shell + Python + dotnet CLI               │
├────────────────────────────────────────────────────────────────────┤
│                                                                    │
│  1.1  Select oldest untracked MAF stable from NuGet                │
│       (clean no-op while any watcher scaffold PR is in flight)     │
│  1.2  Resolve + diff the targeted release-critical package surfaces│
│  1.3  Fetch GitHub release notes from microsoft/agent-framework    │
│  1.4  Run `python3 .github/scripts/update_compat_matrix.py`        │
│          → inserts a new row at the top of compatibility-matrix.md │
│  1.5  Run `python3 .github/scripts/gen_guide_section.py`           │
│          → writes guides/maf-X.Y.Z-migration-guide.md              │
│          → ALSO regenerates guides/maf-current-migration-guide.md  │
│  1.6  Run `registry-extract --diff-file` for validated surfaces   │
│          → scoped drafts, no second network diff, then de-duplicate │
│  1.7  Update .maf-version                                          │
│  1.8  Upload plan/diff+extraction ledgers/diff-*.txt/notes         │
│          as workflow artefacts for reviewer audit                  │
│  1.9  git checkout -b release-watcher/maf-X.Y.Z                    │
│       git commit + git push (per-version branch, NOT main)         │
│  1.10 gh pr create --base main --head release-watcher/maf-X.Y.Z    │
│          (2026-06-28 "C" refactor — an unfilled scaffold turned    │
│           main RED and blocked every downstream PR; Stage 1 now    │
│           always opens a PR instead of committing direct-to-main.  │
│           No longer auto-dispatches Stage 2 below — that's now a   │
│           maintainer's manual, optional next step, targeting the   │
│           branch this PR is on.)                                   │
│                                                                    │
└────────────────────────────────────────────────────────────────────┘
                                  │ manual, optional
                                  ▼
┌────────────────────────────────────────────────────────────────────┐
│  STAGE 2 — AI-fill dispatch (manual, optional)                     │
│  Maintainer runs: gh workflow run maf-ai-fill-todos.yml            │
│                      -f target_version=X.Y.Z                       │
│  .github/workflows/maf-ai-fill-todos.yml                           │
│  Runs on ANOTHER fresh Ubuntu VM                                   │
│  Still no LLM — just creates an issue + assigns Copilot            │
├────────────────────────────────────────────────────────────────────┤
│                                                                    │
│  2.1  Ensure `maf-release` and `ai-fill` labels exist (idempotent) │
│  2.2  Open a GitHub issue:                                         │
│          - Title: "Fill TODOs for MAF X.Y.Z"                       │
│          - Body: full filling prompt (the prompt the maintainer    │
│                  wrote — what to fill, style anchor, constraints)  │
│          - Labels: maf-release,ai-fill                             │
│  2.3  Assign Copilot Coding Agent via GraphQL                      │
│          - bot ID: BOT_kgDOC9w8XQ (stable across all repos)        │
│          - mutation: addAssigneesToAssignable (additive)           │
│          (gh CLI's --assignee doesn't work — REST `/users/Copilot` │
│           doesn't return the bot. GraphQL does.)                   │
│                                                                    │
└────────────────────────────────────────────────────────────────────┘
                                  │ issue assignment
                                  ▼
┌────────────────────────────────────────────────────────────────────┐
│  STAGE 3 — GitHub Copilot Coding Agent                             │
│  Runs on GitHub's own infrastructure (NOT our workflow VM)         │
│  THIS is the LLM/agent part                                        │
├────────────────────────────────────────────────────────────────────┤
│                                                                    │
│  3.1  Reads the issue body (the prompt)                            │
│  3.2  Reads the repo (registry.yaml, matrix, guide, release notes) │
│  3.3  Creates a branch (e.g. copilot/fill-todos-for-maf-1-4-0)     │
│  3.4  Makes file edits — fills the TODOs:                          │
│          - registry: fix_description, example_before/after,        │
│            guide_section (N/A if no parallel in 1.3 guide)         │
│          - compat matrix: real version constraints if release      │
│            notes mention them, else `unknown` + TODO comment       │
│          - per-version guide: Breaking Changes, New Patterns,      │
│            Obsolete APIs, Known Misalignments sections             │
│          - clean up terminal-escape artefacts in diff summary      │
│  3.5  Opens a draft PR titled                                      │
│        "chore: AI-filled TODOs for MAF X.Y.Z"                      │
│        labelled `maf-release,ai-fill`                              │
│                                                                    │
└────────────────────────────────────────────────────────────────────┘

Read the full file on GitHub · 199 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.

  1. 12d ago First seen · 199 lines · 124 tokens per session scan A 6757a90ab537

Subscribe to this mod's changes

maf-release-watcher is a skill published in the GitHub repository joslat/maf-doctor (14 stars, last pushed 25d ago), licensed MIT. It adds 124 tokens to every session and 3,157 once invoked, about $0.0006 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 2026-08-30.