What the board has decided about this committee or office, and the reports it filed, most recent first. Generated from the published minutes.
These categorised views are best effort. They are read from the published minutes by a program, and what it can attribute to a committee depends on how each resolution happens to be worded — so a page may be incomplete, and nothing here overrides the minutes themselves. The raw minutes are the source of truth.
Please note. The board approves the minutes of a meeting at the meeting after it, and they are published then - so the most recent meeting is not usually here. ASF Members can read the draft in the board's own repository until it is.
[raw] the published minutes, which are the record. [main index]
| Date | Action | Decided by |
|---|---|---|
| 2026-06-17 | Appointed Jarek Potiuk | the board, by resolution |
| Date | Decision | Item | Outcome |
|---|---|---|---|
| 2026-06-17 | Establish the Apache Magpie Project | special order | approved |
| 2026-05-20 | Establish the Apache Magpie Project | special order | tabled |
## Description: Apache Magpie is responsible for the creation and maintenance of software related to agent-assisted repository maintainership and development — including issue and pull-request triage, contributor mentoring, agent-drafted remediation, developer-side development-cycle skills, and narrowly-scoped fix-and-merge automation. Magpie is deliberately governance-agnostic: it is built to be adopted by ASF projects and by non-ASF open-source projects alike, with no ASF-only assumption baked into the install path. ## Project Status: Ongoing, high activity (with some low-activity period due to ASF events and vacations). Almost all of this period went into the two things early adopters complained about after 0.1.0: Magpie was hard to install, and once installed it asked for approval too often to be useful. * Marketplace distribution ready. Following industry standards adopted in July/August Magpie can now be installed from a (distributed) agent marketplace instead of committing framework machinery into the adopting repository, and skills can be taken family by family rather than all at once. The primary commands were renamed from adopt/unadopt to install/uninstall, which is what people expected them to be called. This was the largest single piece of work in the period and the direct answer to the adoption friction reported last month. * Permission-prompt reduction is underway, starting with the security family and generalising from there. The aim is for a session to run to a useful stopping point without a human approving each individual step, while keeping the human in the loop for anything that changes state. * No release this period. Apache Magpie 0.1.0 (2026-07-18) is still the only Apache release. The 0.2.0 is being prepared around the installation and permissions work. Though it’s likely we might call it 1.0.0 if it proves to be prime-time-ready - based on feedback of early users. No new issues for the board Thanks to the board for answering our question in the last report. The exploratory conversations with the Python Software Foundation and the Eclipse Foundation continue, and remain exploratory. Nothing formal will be done without engaging VP Marketing & Publicity first. ## Releases * Apache Magpie 0.1.0 — 2026-07-18. Still the only release. Nothing released in this period. * 0.2.0 waits on the marketplace-installation and permission-handling work landing complete rather than on a date. ## Membership Data: Apache Magpie was founded 2026-06-17 (3 months ago) There are currently 25 committers and 22 PMC members in this project. The Committer-to-PMC ratio is roughly 7:6. Community changes, past quarter: - No new PMC members (project graduated recently). - Andre Ahlert was added as committer on 2026-08-17 - Arnav was added as committer on 2026-08-13 - Yeonguk Choo was added as committer on 2026-08-17 ## Project Activity: * 89 pull requests merged in the period, from 20 distinct authors (19 people plus Dependabot) * The security skill family became a complete lifecycle. The threat-model skills used to prepare over 70 ASF PMCs for the foundation's security scanning effort has been made public on 3rd of September https://news.apache.org/foundation/entry/security-scanning-at-foundation-scale were brought into Magpie, so the family now runs from preparing a threat model through triage, CVE allocation, fix and disclosure. Integration with https://dash.security.apache.org/ and a skill that drives an external code scanner are in progress. This matters because the scanning effort has left a lot of PMCs asking how to handle the resulting influx of reports, which is exactly what this family is for. * Security work landed: security-model prepare/verify/update skills; a finder-credit policy with a timing gate; a policy covering reporter-supplied proof-of-concept code; and RFC-AI-0008, which describes handling reports relayed through an intermediary rather than sent by the finder directly (the ASF security team, a disclosure platform, an internal SOC). * Consistency tooling. A set of deterministic checks now runs as commit hooks across specs, evals, docs and the skill/tool tree, catching drift, stale counts and broken cross-references. This is a deliberate response to a real problem: agent-written content moves fast and quietly goes stale, and we would rather catch that mechanically than by reading. * Framework is at 74 skills and 46 tools, distributed as 10 family plugins. * Vendor neutrality remains measured and published at 100%: 10/10 capability contracts, 74/74 skills carrying no vendor lock-in, 22/22 substrate tools harness-neutral. https://magpie.apache.org/#vendor-neutrality * Governance note, for transparency. Last month we reported that approval-before-merge had been turned on for main. That requirement is currently and deliberately lifted while the installation and permissions rework lands, because that work arrives as many small iterations that need testing against a real merged main and a per-PR approval round-trip stalls it for hours at a time. It is time-boxed and documented in .asf.yaml, not a governance change: CI status checks, review-thread resolution and linear history are all still gate main. Effectively, CTR (Commit-Then-Review) mode is used for now rather than RTC (Review-Than-Commit). It goes back on when this work is done (unless we decide otherwise). * Submitted proposal for Devroom at FOSDEM: “Agentic Automation for OSS Maintainers” ## Community Health: * Three committers added — Arnav Purohit (2026-08-13), André Ahlert and Yeonguk Choo (both 2026-08-17). These are the project's first committer additions since establishment. No PMC additions or removals. 25 committers, 22 PMC members. * Contribution is still concentrated. Of the 89 PRs merged this period, the Chair authored 43. The PMC is clear that this is the main community-health risk. The three new committers and the growing tail of contributors are the counterweight; it will take more than one period to show. * The other 19 authors were mostly people outside the founding PMC, so the newcomer path is working even though the commit share is not yet balanced. The good-first-issue programme is now at 111 curated issues, 81 closed. * Community Over Code Asia - the Chair gave security keynote and a Magpie “community” talk at the Community Over Code Asia in Beijing in August - the reception was very warm and a lot of discussions at the event resulted in reaffirming that security issue management and triage are a common pain point among the community, but that easiness of use and low friction is a key for Magpie’s popularity. * Someone is using it on a real problem without being asked to. A community member has been running Magpie against Apache Iceberg (iceberg-go) to work through security scanner findings, and has been recommending it to the rest of that PMC. That is the adoption shape we want — a maintainer with a backlog, not a demo - first time where our Marketing (speaking to others about us) turns into PR (others talking about us). * Community Over Code Glasgow: A hackathon slot is reserved for Tuesday 12 September to get PMCs onto the security skill family, alongside the security team's office hours. Four community members have already volunteered to help run it. The intent is to convert "people don't know Magpie exists or that it solves their problem" into projects actually using it. * Skill quality is being addressed by hand. The PMC agreed that most skills, having been written by agents, are longer and less predictable than skills written by people. Before 0.2.0 we are going through the important ones and rewriting them in plain human English — shorter, clearer and inside the context budget. The agreed method is to use a skill first and then rewrite it from that experience, which doubles as onboarding for PMC members who have not yet used the framework closely. * Still open from the last two reports: ASF Slack requires guest invites for non-ASF participants, which works against being open to adopters outside the foundation. The PMC has not yet chosen between Matrix and Discord for an open channel.
## Description:
Apache Magpie is responsible for the creation and maintenance of software
related to agent-assisted repository maintainership and development —
including issue and pull-request triage, contributor mentoring, agent-drafted
remediation, developer-side development-cycle skills, and narrowly-scoped
fix-and-merge automation. Magpie is deliberately governance-agnostic: it is
built to be adopted by ASF projects and by non-ASF open-source projects alike,
with no ASF-only assumption baked into the install path.
## Project Status:
- First committer votes succeded — three candidates, our first since
establishment. These are committer votes only; whether committers should
automatically join the PMC remains an open question the PMC is discussing
separately. We are proceeding with inviting and formalising it.
- PMC/Committers: 22 members. Chair: Jarek Potiuk. No additions or removals in
the period.
- Apache Magpie 0.1.0 — 2026-07-18. Our first release.
- Cut via a hybrid path: the Apache Trusted Releases (ATR) platform with the
traditional manual process as fallback. We used the fallback when ATR hit
blocking issues, and fed that experience back to the ASF Tooling team.
- Two release candidates were needed; rc1 was withdrawn after review
feedback. We consider that the process working rather than a problem.
- The release was itself driven by Magpie's own release skills — dogfooding
that surfaced real defects we have since fixed.
No new issues. As noted last month, the PMC is exploring relationships with
other open-source organisations. These remain exploratory discussions only;
nothing formal will be done without first engaging VP Marketing & Publicity.
The initial discussions had shown that while there is interest - the
installation approach for Python Software Foundation use and other non-ASF
projects might be a blocker. Eclipse is more interested in having Enclave
embedded as part of the sandboxing. Both approaches are in the discussion
phase.
Thanks for clarifying our question from last month.
## Membership Data:
Apache Magpie was founded 2026-06-17 (2 months ago) There are currently 22
committers and 22 PMC members in this project. The Committer-to-PMC ratio is
1:1.
Community changes, past quarter:
- No new PMC members (project became TLP recently).
- Three new commmitters are already accepted, formalising the process was a
little impaired due to PMC chair attending Community Over Code Asia.
## Project Activity:
Ongoing, medium activity. The project has shipped its first Apache release and
is now moving from bootstrapping into normal operation.
Discussion at Community Over Code Asia - where three PMC members (Jarek, Craig
and Justin) were speaking had shown that Magpie addresses the very problems
many of the PMCs in Asia are also struggling with. Magpie was mentioned
multiple times at Keynotes and other talks (alongside RAI initiative as well).
The discussions (also earlier ones) allowed to identify two main bottlenecks
for adoption:
* Installation process
* Too many interruptions due to security / permissions
Luckily those are common industry problems and industry is converging on
solutions: Marketplace and Plugin standards have just been released by Open AI
and Anthropic, and permission models for CLI is changing to "auto-approve by
default". We are going to leverage both in the next version - while we are
going to build addional security layer, which will maintain current security
level while avoiding too frequent interruptions.
- 162 pull requests merged in the period, across roughly 37 contributors.
- Governance tightened. Approval-before-merge is now enforced by branch
protection; the project no longer self-merges without review. This was a
deliberate slowdown following the release.
- 0.2.0 focus is adoption friction.
- Trusted external skill sources (RFC-AI-0006) landed, allowing skills to be
pulled from trusted external repositories — the mechanism the PSF
conversation depends on.
- Vendor neutrality remains automatically measured and published at 100%
across capability contracts, skills, and agent harnesses:
https://magpie.apache.org/#vendor-neutrality
- Framework now at 70 skills and 45 tools.
- A read-only repo-health audit family shipped: Actions workflow security,
dependency vulnerabilities, licence and NOTICE compliance, and flaky-test
patterns. These skills observe and report only; they make no changes,
consistent with the project's human-in-the-loop discipline.
## Community Health:
Healthy and broadening, with some caveats.
Contribution is still concentrated: Jarek and Justin account for a large share
of commits. But the tail has grown — several contributors now show sustained,
substantive activity, and the three committer votes now running are the first
step in converting that into shared responsibility. Enforcing
review-before-merge was the other half of that deliberately, since it forces
work through more than one pair of hands.
Interest continues from outside the ASF, including a contributor with
professional AI-security and privacy background who approached the list
directly.
The Chair was presenting a Magpie talk and the security keynote at Community
Over Code Asia, We are also discussing an ASF blog post intended to carry
perspectives from adopters rather than only from the PMC. Further events that
are coming where we can get further feedback are Airflow Summit and Community
Over Code Glasgow.
Ongoing community discussion: ASF Slack requires guest invites for non-ASF
participants, which conflicts with our goal of being open to adopters outside
the foundation. The PMC is deciding between Matrix and Discord for an open
channel.
Onboarding investment has improved. An 11-lesson AI tutor curriculum now ships
in the repository, giving new contributors and adopting maintainers a
structured path into the framework.
Also for the upcoming release (0.2.0) we are reviewing all the SKILLs, docs
onboarding docs for consistency. While Agentic driven development is great to
move fast, it also breaks some consistency, miscount things and hallucinates
sometimes, we are not only fixing the inconsistencies but adding harnesses to
avoid those in the future.
A sustained good first issue programme has raised around 60 curated newcomer
issues, with 45 closed and roughly 25 distinct external contributors, with
several now repeat contributors.
## Description: Apache Magpie is responsible for the creation and maintenance of software related to agent-assisted repository maintainership and development — including issue and pull-request triage, contributor mentoring, agent-drafted remediation, developer-side development-cycle skills, and narrowly-scoped fix-and-merge automation. Magpie is deliberately governance-agnostic: it is built to be adopted by ASF projects and by non-ASF open-source projects alike, with no ASF-only assumption baked into the install path. ## Project Status: ### Status: New Top-Level Project, established directly by Board resolution (not via the Incubator). Status: ongoing, high activity — in the bootstrapping phase. The project is standing up its infrastructure and framework, onboarding its founding PMC, and preparing its first Apache release. No issues require Board attention at this time. * Established: 2026-06-17 (name confirmed via PODLINGSEARCH during the 8–12 May 2026 naming discussion). * PMC: 22 members at establishment. * Chair: Jarek Potiuk. * No committers or PMC members have been added or removed since establishment (project is newly formed). * Diversity: the founding PMC spans a broad set of ASF projects (Airflow, Arrow, Iceberg, Polaris, Groovy, Grails, Superset, Beam, Pulsar, Log4j/Logging Services, Mahout, Doris, Geode, Accumulo, HBase, Lucene/Solr, Tomcat, ComDev, Incubator, Security, Training) and multiple organisational affiliations, with no single employer holding a majority — a deliberate design goal given the contributor-data surface the project touches. * There are first new adopters discussing Magpie Adoption (Apache JMeter in the ASF, but also we have a discussion with Eclipse Foundation Leadership and we have active discussions with Python Software Foundation - and in the week of 13-19 of July at Europython we are planning to work with release managers and triagers of Python to help them to adopt Magpie and possibly expose their SKILLs via trusted 3rd-party mechanism of Magpie. ### Issues for the board: The announcement of Apache Magpie triggered immediate adoption interest aligning perfectly with our core mission. This momentum extends well beyond the ASF; we are seeing active usage and contributions from independent maintainers, alongside formal interest from both the Python Software Foundation and the Eclipse Foundation. This "grass-roots" pull comes from maintainers seeking to reclaim their time through Magpie's automation. While this outreach intersects with broader initiatives like RAI, the PMC views this as project-specific work addressing universal OSS maintainership challenges. We intend to continue these cross-organizational collaborations—engaging Marketing & Publicity should they evolve into formal public partnerships—to ensure Magpie remains a governance-agnostic solution for the entire ecosystem. Question to the board: Does the Board have any objections to the PMC pursuing these relationships with other open-source organizations, or do you have any other expectations or constraints regarding them? ### Releases * No releases yet. The project has made no Apache release to date * The first release, Apache Magpie 0.1.0, is being run https://lists.apache.org/thread/4cxf84q5vwxd85dj5yx6vhj2bzr46111. * Magpie intended to cut its releases through the Apache Trusted Releases (ATR) platform (ASF Tooling) with traditional manual release process as fallback - dogfooding its own release process, which turned out to be good idea as we could fallback to manual process when ATR hit some un-workaroundable issues. * Releasing Magpie using release SKILS in Magpie is a fantastic way to dog-food what we are doing and shows the true value of Magpie - the whole release process was generally following “release Magpie” Agentic command - with conversational interface guiding release manager - including automated installation of dependencies, generating keys etc. It was as easy as following the instructions of the agent. ## Membership Data: Apache Magpie was founded 2026-06-17 (22 days ago) There are currently 22 committers and 22 PMC members in this project. The Committer-to-PMC ratio is 1:1. Community changes, past quarter: - No new PMC members (project graduated recently). - No new committers were added. ## Project Activity: Since establishment the project has focused on infrastructure and framework build-out: * Repository and framework. github.com/apache/magpie is stood up with the skills framework, contributor documentation, and CI. The framework ships as reusable "skills" a project adopts via a snapshot-based adoption mechanism, with per-adopter and per-user configuration layers and a strict human-in-the-loop state-change discipline (the agent drafts; the human acts). * Magpie Website reflecting Maintainer focus and agreed Taxonomy Targeted to maintainers who are main audience for Magpie has been published - promotion of the “Give maintainers time back” theme is happening in various media to attract contributors: https://magpie.apache.org/ * Skill families in place across the project's agentic modes: security-issue handling (intake → triage → CVE allocation → fix → disclosure), release management (full lifecycle, both svnpubsub and ATR backends), PR management (triage, review, quick-merge, stats), issue management (triage, deduplication, reproduction, stale sweeps), contributor/committer (nomination, activity, onboarding), and audit (CI-runner, dependency, license-compliance, flaky-test). * Pilots. Onboarding friendly pilot projects — an ASF PMC running the full security-issue flow (Apache Airflow, Apache Groovy, Apache Burr, Apache JMeter) * Trusted-release collaboration. Working with the ASF Tooling team on the ATR platform, adopting it for Magpie's own first release as an early real-world test. * Vendor neutrality - discussions started on vendor neutrality of the project being aspirational vs. achieved - this resulted in implementing components and defining contracts that allow Magpie to claim Vendor neutrality status. Vendor neutrality is automatically assessed at https://magpie.apache.org/#vendor-neutrality The current state: 10/10 Capability contracts 100% work across 2+ vendors 69/69 Skills carry no vendor lock-in 22/22 Agent harness 100% harness-neutral * Organization agnosticism - discussions started with Python Software Foundation and Eclipse Foundation on a way for Magpie to become a way to also distribute other organization skills via trusted skills https://magpie.apache.org/#organisation-agnostic ## Community Health: Healthy for a project at this stage. Development activity is high (framework and skill build-out proceeding steadily); the PMC is engaged and drawn from experienced ASF contributors with a bias toward developer-experience and community-development work. Mailing lists (dev@, private@, commits@) and GitHub Issues are active for coordination. The immediate community-building priorities are landing the first release, converting pilot usage into external adopters, and beginning to grow committership beyond the founding grant. We have a great deal of contributions - mainly from involved founding team members. Jarek Potiuk, Justin McLean - who both happen to have more time to dedicate for Magpie - are most active contributors. However we make an effort to get more feedback from other founding team members in the decision making process and project daily operations - we also have founding team members contributing and new, individual contributors who contribute occasionally. We have active discussion on Slack #magpie channel created on ASF slack, and we are discussing about adding extra channels (Discord) for ad-hoc discussions and ASF slack adds friction in adding non-ASF community members - that have to be personally invited and approved as guests, we want to make sure we are as open as inclusive as possible to all kinds of organizations and individuals - in- and out- of ASF. The “Mischief Makers” is a term we seem to be coining to name community members - which seems a pretty natural choice. Thanks to Agentic assistance it progresses very fast - we have now 10 contracts, 5 substrates, 69 skills, 45 tools, 9 skill families that are ready to serve Maintainers all-around globe.