Repository navigation
Promote 0.8.0: campaign-owned report defaults - #72
Closed
bioinfornatics wants to merge 144 commits into
Closed
bioinfornatics wants to merge 144 commits into
bioinfornatics wants to merge 144 commits into
Conversation
…tion docs: bilingual pedagogical Zensical site and GitHub Pages
Integrate remaining documentation and ECC commits before branch cleanup
…ming Use neutral capability research naming
…iming Improve Skill Creator selection and action-local instruction flow
…-hook-targets fix(plugin): safe discovery and source-pinned Goose hooks
… [ap-t06 ap-7r2 ap-yiy ap-2wc]
…chestration Session-owned evaluation orchestration and HTML reports
Rebuild creator scenarios from active instruction journeys, preserve historical resources, and add trusted retained preparation with authenticated bound tasks. Keep global XDG data registry separate from per-attempt synthetic fixtures. Verified: app build; npm test (254 pass, 1 optional skip); typecheck; lint; source inventory; runtime drift; distribution (13); production (15). Native provider effectiveness remains unverified; consultation enabled, nested resume/grading pending. Beads: ap-9ato, ap-wljl.
CI failed before tests because the new app regression was absent from the exhaustive partition. Verified ci-tests plan and 28 CI/inert/reproducibility checks; no exclusions or assertion weakening.
Restore explicit fail-closed layout and manifest-preservation requirements. Update source-contract assertions to current manual-review wording without weakening event, activation or permission boundaries. Hook source checks 3/3 pass; local remainder stops at missing explicit pinned Goose, CI supplies it.
…retained-preparation Align creator Skills and add persistent retained trial preparation
Preserve frozen inputs and private-HOME authored Skills for grading; bounded neutral evidence reader and retained provenance. Simplify evaluation cases and remove retired creator/native tooling and obsolete fixture corpora. Add host-specific hook inventories. Verified app build/typecheck/lint, generated runtime drift, full root suite with explicit Goose 1.57.0, documentation source links and inventory. Beads: ap-9ato, ap-9ato.2, ap-9ato.3, ap-9ato.6, ap-9ato.7. No model campaign executed.
Restore the app compilation prerequisite in each fresh CI shard checkout. Preserve the real compiled MCP reader guarantee and assert prepare/build/test ordering. Verified clean isolated shard 4 (71/71), CI regressions (17/17), inventory and diff checks; independent review accepted. Beads: ap-9ato.8.
…e-and-cleanup Fix evaluation evidence and simplify creator suites
Emit a private-data-free stderr heartbeat every 10 seconds for evaluation, resume, grading and regrading; -q/--quiet disables only heartbeat. Keep stdout results intact and clean up unreferenced timers. Document detection, same-process waiting, buffered output limits and final-status checks in active Skills. Include generated runtime and declared source/test inventories. Verified: app 307 pass, 1 optional skip; build, typecheck, lint, runtime drift and inventory; distribution 13 pass, production 15 pass. Independent reviews 20261107_59 (implementation) and 20261107_60 (instructions) accepted local scope. No model calls or live installations. Beads: ap-a00a, ap-k261.
Reuse the production projection and generic evaluator without repository coupling. Preserve historical fixtures, move active neutral inputs under evals, and verify candidate inventory and treatment integrity. Fix Markdown fragment handling in distribution file checks. Task: ap-ir5d. Verified: distribution 14/14, candidate 4/4, source inventory and diff checks; earlier production 15/15 and CI/reproducibility 25/25. Separate local reviews accepted. No model execution or publication.
Show formal grade summaries before conversations and collapse JSONL/log diagnostics without altering embedded bytes or downloads. Select a unique available condition on case/repetition navigation; keep Compare and explicit missing-arm diagnostics without cross-repetition pairing. Verification: Node 22.22.2 targeted report tests passed 4/4, including two Chromium browser tests (zero skips); independent local reviews accepted. No runtime build or model execution. Beads: ap-a5mk, ap-09kc
Retain production and evaluation trees from a single projection. Bind canonical candidate artifacts to exact source/run/attempt, authenticate archive digest and restore bounded data with executable intent. Promote retained bytes without rebuilding and verify staged blobs before publication. Task: ap-6h9m. Verified on Node 22.22.2: 55 targeted tests pass including real CLI/local bare remote success and corruption refusal; source inventory and diff checks pass. Goose install and GitHub approval services are explicit integration-test doubles. Live workflow/artifact rollout and behavioral model acceptance remain unverified.
…romotion Promote frozen evaluation candidates without rebuilding production
Replace custom Markdown parsing with pinned offline browser libraries, strict sanitization, safe links, and GFM tables. Preserve exact transcript bytes and embed libraries/licenses once per standalone report. Fix placeholder collisions and reject encoded asset-path aliases. Include contained-library integrity and browser safety regressions. Verified staged snapshot in an isolated copy: Chromium report tests 5/5 (zero skips), distribution 14/14, production 15/15; staged diff check passed. Working-tree source inventory passed. Independent scoped review accepted after corrections. Beads: ap-lc2p
Default ordinary Skill and Plugin evaluation requests to preparation, paired trials, initial grading, explicit export and new review.html. Resolve execution authority once and preserve explicit partial workflows, failures and no-retry boundaries. Add neutral planning and preparation-only scenarios and static distribution routing guards. Local distribution/production tests passed; no paid behavioral campaign claimed. Beads: ap-m892
Update exact suite cardinality to 18 and require both added Skill Creator scenarios with their neutral declared input. Preserve all production/integrity checks. Reproduced stale 16-scenario oracle; targeted candidate/distribution/production 34/34 passed; independent scoped review accepted. Beads: ap-b73h
Complete evaluation DX and safe offline report rendering
Add frozen all-trial overview, explicit source identity, model-separated score partitions, compact cards and persistent feedback. Accept documented Goose thinking auxiliaries without weakening verdict validation; retain safe structural diagnostics. Verified full root/app suite with Node 22.22.2, Goose 1.57.0 and real report browser tests; typecheck, lint, runtime drift and inventory pass. Release documentation and installed overview inventory gates verified. Beads: ap-bw5x ap-6hnu ap-e263 ap-2d4c ap-qfsf ap-92xq ap-6aka ap-6avo.
…w-0.6.1 Release 0.6.1: evaluation overview and safe judge diagnostics
Separate validated verdict privacy from raw diagnostic parsing; add bounded structural tool evidence and condition-aware report cards. Preserve frozen grades and distinguish baseline absence from future consultation claims. Local verification: app 343 passed/1 optional skip; typecheck/lint; 9 Chromium report tests; distribution/production 15 each; runtime generation/drift, inventory and documentation source checks. Independent release-diff review accepted. Beads: ap-sb8p, ap-3mwp, ap-x393.
…lity-0.6.2 Release 0.6.2: reliable grading and partial comparisons
Implement ap-vh5g, ap-yy6y and ap-h8f6. All CLI campaign roots retain Hive layout; frozen configured provider/model identities survive default changes and resume, with session observations kept separate. Preserve legacy evidence and omit unsolicited optional CLI overrides in Skills. Breaking pre-1.0 path contract requires minor bump from main 0.6.2. Verified local app 368 pass/1 optional upstream skip, types/lint, source inventory, deterministic rebuild/runtime drift, distribution/production 15 each, independent source and release-companion reviews. No paid-model/backend compatibility claim.
…del-0.7.0 feat: 0.7.0 Hive campaigns and frozen candidate models
Commit the user-approved root AGENTS.md policy. Require a strictly increasing distribution version, explicit pre-1.0 breaking-change minor bumps, frozen candidate publication, and separately authorized tagging. ap-2b9s. Verified source documentation parity/links and source inventory; git diff --check passed. Beads journal excluded.
…policy docs: distribution versioning policy
Deliver ap-mn3w and default-preserving nested command guidance (ap-9hby, ap-xyl7). Managed exports and HTML use recorded campaign paths, authenticated replacement and explicit commit/cleanup boundaries; evidence and grades remain immutable. Verified: app 376 pass/1 optional skip, helper 12 pass/5 browser skips, typecheck/lint/build/runtime drift, distribution15/production15, source inventory. Independent local reviews 20261109_29 and release companion 20261109_30. No paid provider or live installation qualification. Delivery tracked by ap-qtwt.
…faults feat(reporting): campaign-owned default report views (0.8.0)
Contributor
|
Published generated runtime distribution from 49bba1f. Closed without merging authoring files into main. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Owner explicitly authorized generated main promotion on 2026-11-09. Source49bba1f45379c1bcb1339821e239e09174d69140 integrates PR71. Version0.8.0 is greater than current published0.7.0. Await canonical exact-source push/develop seven-job verification; publish only its retained frozen production artifact. This request is review/authorization metadata, not a source merge: close without merging after verified promotion. Ordinary source/distribution conflicts are expected. Local feature and release reviews accepted; no paid-provider/live-install/browser efficacy claim. Delivery ap-qtwt.