DottaandPaperclip
0e15c92073
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Agents use installed skills and exact version pins during execution. > - Managed skills still depended on mutable directories on the web server. > - A replacement server could lose acknowledged edits or remove an unassigned skill whose directory was absent. > - This pull request makes immutable versions the authority for managed content and installed repository reads. > - The server can rebuild delivery files from durable content after replacement. > - This establishes skills readiness for ephemeral web servers. General server draining and workspace recovery remain separate work. ## Linked Issues or Issue Description **Subsystem affected** Skills services, shared contracts, database receipts, Skill Studio, and runtime delivery. **Problem or motivation** A fresh web server must discover, read, edit, and deliver protected installed skills using the same database, object storage, secrets, and supported application artifacts. It must not need its predecessor's files or process memory. **Proposed solution** Use `managed` ownership with a null source locator. Read installed content by immutable version. Prepare packages before a short publication transaction. Persist execution selections before materialization. Recover exact pinned runtime assets and reject deletion while active or recovering pins exist. **Alternatives considered** Reconstructing a mutable installed directory preserves the old source-of-truth ambiguity. Permanent dual writes create two competing authorities. This change uses one managed model after a bounded legacy conversion. **Roadmap alignment** This improves the existing Skills Manager, Studio, version history, and Cloud execution capabilities. The maintainer supplied the implementation contract. This does not add a separate skills product or deploy fleet changes. Related work: #18082 supplies immutable package storage. #17273 addresses edits outside the API for linked local skills; this PR keeps that source category separate. #2648 is a closed earlier ephemeral-filesystem proposal. ## What Changed - Add managed ownership and focused content publication and execution materialization modules. Keep the archive format and package storage tables. - Publish managed create, save, delete, fork, rename, restore, and catalog content with version checks. Add durable create receipts. Preserve file-save receipt responses. - Return file version identities. Keep Studio drafts after a conflict and provide an explicit reload action. A delayed reload cannot overwrite newer typing or navigation. - Persist exact runtime selections, including empty selections. Rebuild pinned assets and verify their admitted digests on a replacement server. - Preserve source access on forks and disconnected copies. Keep the existing workspace policy category for managed skills. - Add bounded conversion, eligibility, and independent verification phases. Load persisted Cloud configuration in the standalone CLI without seeding or changing it. Reject partial current versions against installed inventories; recover only complete saved content or the exact source. Document storage migration blockers and the compatible rollback floor. - Add separate-process PostgreSQL/S3 qualification and update domain, retention, policy, and UI regressions. - Pass the plugin worker manager into Studio execution, admit restricted `skill_test` mode through native Runner contracts, and preserve admitted scope across native and REST tools. Native tests cannot create ordinary tasks, browse company data, or use external tools. - Select the exact durable Studio revision even when unassigned or Beta skills is off. Persist Studio output and audit atomically with native task completion, preserving the result on replay. ## Verification - Final revision `13e8af480b88ba6fea5737bac4bd81104d604792` passed full CI: all 49 workflow jobs completed successfully ([run 41127396098](https://github.com/paperclipai/paperclip/actions/runs/41127396098)). This includes typecheck, test suites, Runner checks, browser checks and build. Greptile review 15 is 5/5 on that same revision, with zero unresolved threads. The PR is ready for review. - Local workspace typecheck and build passed after integrating master. Focused checks passed: 33 execution/completion cases, 52 Runner HTTP cases, 16 managed-content cases, 17 Studio cases, the new real-process CLI regression, map checks and UI token gates. The CLI regression proves inspection does not seed missing Cloud configuration. Server typecheck passed after that fix. - The independent-process PostgreSQL/S3 matrix passed through normal HTTP skill routes. It covers A → B → C replacement, lost commit responses, upload crashes, concurrent saves, binary contents and executable delivery. Metadata listing and warm selected delivery retain zero object downloads. Local timing measurements are recorded; they do not establish hosted latency. - Hosted A → B → C passed on the owned disposable test tenant with external Postgres, S3 and a real remote Paperclip Computer. Predecessor temporary homes were absent. C read and executed B's acknowledged managed save. New runs used repository v1 while v2 was installed. Hashes matched an independently checked exact upstream commit. - Compatible image rollback to `6518b323` passed. Studio retained the unassigned fork's C save. A fresh remote run executed the newer managed content and historical repository pin. The actual shell event confirmed expected hashes, stdout, exit 0 and mode 0555. - Runtime candidate `9200e3fcb4` deployed successfully ([Cloud run 41124147876](https://github.com/paperclipai/paperclip-cloud/actions/runs/41124147876)). The final Studio test delivered only the unassigned fork's exact v3, executed its script and persisted a successful output snapshot. Independent command-event inspection confirmed all three hashes, stdout, exit 0 and mode 0555. All 14 installations were version-backed. Independent verification checked all 25 versions: 23 remote objects, two bundled database snapshots, zero blockers. - The deployed image is `9200e3fcb4`. Later master integration, the client-only delayed-reload draft guard and the CLI bootstrap have separate automated proof. The CLI bootstrap was also exercised against the tenant's stored configuration and storage through the same read-only services. This PR does not claim that the deployed image is the final PR revision. - Live Studio checks exposed worker wiring, admission, access scope, exact selection and completion-publication gaps. They are fixed. Failed attempts remain in the evidence. The full local suite exceeded 80 minutes under host contention and was stopped. Later combined/isolated local attempts also timed out; they are not counted as passes. Full CI covers the integrated suites. Earlier CI found an obsolete restart-scope fixture, which was corrected and passed all 52 HTTP cases. - Reproduction and migration: [`doc/skill-ephemeral-web-servers.md`](doc/skill-ephemeral-web-servers.md). Source revisions, deployments, versions, run IDs, hashes and failure history: [`doc/plans/2026-10-30-ephemeral-web-skills.md`](doc/plans/2026-10-30-ephemeral-web-skills.md). ## Risks - Legacy conversion requires stopped admission and retired filesystem-authoritative writers. Pre-conversion binaries are not supported rollback targets afterward. - Cloud readiness requires remote object storage. Local provider receipts and provider mismatches remain blockers until objects are copied and verified. - Linked local sources retain filesystem semantics and are outside the cloud guarantee. - Cold reads still fetch the existing archive format. Warm runtime preparation retains the verified-cache path. - Bundled reconciliation preserves an installed head. Existing release snapshots remain selectable. An older server must not replace a newer head with its shipped files. - No production volumes, fleet defaults, or general deployment coordination change in this PR. ## Model Used OpenAI Codex, GPT-6 family, with reasoning, repository tools, code execution, and browser use. The exact backend model ID and context-window size are not exposed in this session. No subagents were used. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing>
12 KiB
navigation
| navigation | ||||||
|---|---|---|---|---|---|---|
|
Skills and Skill Studio
Operators discover or import skills, author and test them in Skill Studio, manage source versions, and decide which agents may use them.
Feature ID: skills.
Implementation: skill library, studio, sources, agent skills.
Agent access: companion.
Sub-features
discovery:skills/discovery— browse installed/discover views and import a skill or source.authoring:skills/authoring— create/edit a writable skill or fork a read-only source.versions:skills/versions— inspect revisions, source updates, diffs, and selected releases.policy:skills/policy— apply agent enablement and skill usage policy independently from installation.testing:skills/testing— run a bounded Studio test and inspect its result and interactions.
How to get to it (user POV)
library-sources
Open /skills, Discover, or /skills/sources; project import is available through the library’s import flow. In Streamlined UI, Installed, Discover and My Skills are route links styled as tabs beneath the toolbar, with no secondary navigation rail. Sources stays separate under New → Manage sources. Legacy UI retains the Skills sidebar.
studio
Use New skill or /skills/studio/:skillId; writable copies and source-backed originals differ. In Streamlined UI, /skills/studio uses the shared Skills toolbar and card collection with My Skills active and Recently updated sorting; opening a card enters Studio.
Direct authoring starts at /skills/studio/new: fill Name and Slug, then
Create skill. Authoring/save/reload requires no model credential; running a
Studio test is separate.
agent-enablement
Use skill policy/agent selection in the library or an agent’s Skills tab.
Driving it with Playwright and the Paperclip CLI
The explicit create-skill-studio Product journey covers skills/authoring: native creation of an owned company skill, its task activity card, a Studio edit, reload, and task-card readback. It does not establish policy, source-update, or Studio-test coverage. A deployed target requires its own live result.
Preconditions: follow the baseline. Use a harmless skill fixture and test agent with skill support. Record whether the source is local, bundled, project, or remote.
library-sources
Automated: library UI and source service cover their respective layers.
Manual: Install a fixture, locate it in Installed, inspect its source and files, and reload. Import a source update and compare versions before accepting it. Check a source requiring unavailable repository access and verify the failure is explicit.
studio
Automated: Studio covers UI behavior; model-driven skill testing remains a separate manual result.
Manual: Create or fork a disposable skill, edit instructions, save, and reload. Run a bounded test, respond to any question, and inspect the output. Verify the saved version matches what was tested and failed saves retain the draft.
For managed skills, open the same file in two Studio sessions. Save through one, then save the other's stale draft. It must receive a conflict, retain its text, and offer Load latest saved file with discard confirmation. File reads and mutation results identify the committed version. Managed content has no server source-directory repair flow. The domain regression and process qualification cover fresh-home recovery; remote Runner staging remains a separate live proof.
Version history lists file metadata without downloading archives. Opening a diff fetches only its selected versions; restoring fetches the chosen version. The library UI regression checks deferred loads; the package regression checks that slow archive verification cannot hold up a warm agent preparation.
agent-enablement
Automated: policy service and agent skills state cover policy/state contracts.
The version-cache regression checks that listing assignments does not create repository caches, validate every support file, or extend cache use. Run preparation still restores and repairs the selected version. To check attachment, use a repository skill with a cold cache, add it to an agent, then reload the skill's Agents tab and the agent's Skills tab. Keep another repository skill set to Latest during this check. With beta version pins disabled, adding the new skill must preserve that Latest selection and save without a beta-pin error. The HTTP regression checks this round trip and retention of explicit historical pins.
Manual: Enable the skill for one test agent and leave another without access. Run a small task on each, inspect the available skill context, then change the release/policy and verify the next eligible run uses the intended version.
A Paperclip Runner run with no selected skills must proceed while another run holds the company skill-cache lock. A saved empty selection remains empty after later assignments. Selected skills and the legacy operational skill retain cache exclusion through version pinning. The configuration regression uses a real PostgreSQL lock; live provider latency remains a separate check.
Concurrent selected-skill preparations share one in-flight company inventory refresh before cache locking. Each run retains its own selection and durable version pins. The concurrency regression checks shared work and separate actor access; the cache tests check that cleanup cannot remove a snapshot before the run stores its pins.
The package storage regression covers external snapshots, automatic legacy conversion and exact-commit recovery. A warm selected-repository check must perform no object reads or repository-content SQL. Source/library reconciliation remains available for discovery and mutable sources. The native context regression checks that warm asset packaging reuses verified files, repairs missing/corrupt copies, and notices source edits. Verify both writer-switch states during rollout; see package storage.
Gotchas
-
The release picker requires
enableBetaSkills. Do not invent an Instructions accessibility label for the Studio body editor; inspect its current handles. -
Installed does not mean enabled for every agent.
-
Read-only GitHub/bundled skills may require a writable copy rather than an in-place edit.
-
Studio interaction coverage is in questions and approvals; a passing mock is not a live harness test.
-
Old version disk caches may expire; saved versions remain available. Verify a pinned version is restored through normal execution with upstream unavailable. Current and in-use versions stay materialized. See cleanup controls and the database-backed cache tests.
Project repository source behavior
New GitHub project repositories import skills automatically. Skills → Sources lists accessible legacy repositories with Import skills, shows each saved repository once with its access scope but without project names, and groups matching repository/ref workspaces. Additional legacy imports for an existing source live in its More menu. Import progress appears in the source metadata; Refresh and Automatic updates are also available. Verify future additions, retained exclusions, error recovery and reload persistence. Private-only sources must be absent for unrelated people/agents; access through either originating project is sufficient.
For an oversized repository, preview and import a skill that links to shared files outside its folder. Its saved version should include those files under the original repository paths. Missing references remain visible warnings and must not disable an otherwise valid skill's selection. Verify an existing source with cached dependency errors recovers on refresh without requiring a new upstream commit. A refresh that cannot preserve installed shared files retains the prior version.
Storybook/browser coverage uses scoped fixtures. Live GitHub-to-routine execution must be verified separately; inspect resolved version IDs across an upstream update.
Connection repair during an update
A missing tracked branch is a branch repair, not a credential failure. On an independent source, choose a new branch from the searchable list or enter its name. The repository's default branch is marked and included even outside the first 100 listed branches. Loading or a failed list request does not prevent manual entry; saving validates the requested ref. Installed identities, selections and version pins remain. Project-tracked branches are changed in project settings. A saved connection is validated against both the repository and tracked revision before repair succeeds.
Open Choose connection while a source is being refreshed. Select an authorized account and save. If the source changes, the dialog must retain that selection, load the current revision, and allow a second save without reopening. The project repository browser suite covers this recovery with simulated upstream data.
Storybook verification
testing
Capability ID: skills/testing.
Scope: ui Storybook: Studio test inputs/configuration, running, question/approval, result and failed test states. Production: Run bounded test of actual saved version with correct selected account/environment; inspect persisted test detail and eligible interaction response. Remote Runner selection accepts the Studio test work mode and preserves its restricted scope; live qualification must cover the saved version delivered to the selected Computer, including a skill with no permanent agent assignment. The execution configuration regression checks that Studio's historical version replaces ordinary assignments without changing the agent's configuration. Use an isolated Runner home for this journey. Sources: Source: SkillStudio.tsx, companySkills.ts. Tests: SkillStudio.test.tsx, company-skill-test-runs-service.test.ts
Documentation scope: Skills, Skill Studio, and teams
Main library/Studio/source/policy and CLI team-install paths are mapped. Full source-provider conflicts, harness sync, and remote package variants remain; catalog UI is not a current top-level route.
Installed skill updates and assignments can drive opt-in plugin runtime synchronization. Version pins remain effective. See contract and database coverage.