v0.7.0 — Blind controlled studies, absorbed reporting standards, de-AI tone pass

Highlights

Evidence: two blind controlled studies published

• evals/runs/controlled/codex-20260825/ and codex-20260826/ — each 28 held-out cases × 2 conditions × 3 fresh-context seeds = 168 real Codex executions, 84 randomized A/B pairs blind-rated by two independent non-OpenAI model raters, ratings frozen before unblinding, paired bootstrap (10,000 resamples) against preregistered gates.
• Second run passes the primary quality gate for the first time: composite +0.315, 95% CI [0.110, 0.579] (threshold +0.3); framework ahead on 6/7 dimensions, concision at parity, win rate 33.3% → 45.2%, zero framework responses leaking a local path (45 of 46 critical-error records on the baseline side).
• Efficiency gates still fail structurally: 7.9x median output-token overhead (p90 26.7x) against a 1.15x/1.3x budget; semantic density per 1,000 tokens 1.9 vs 19.9. The controller keeps effectiveness_claim_eligible: false and the claim status insufficient_evidence. No readability or efficiency gain is claimed. Both READMEs' Evidence status sections now report this instead of only the one-case pilot.

Contract fixes motivated by the first study

• The final response must contain the report itself — never a saved-file path, pointer, or scratch path — in every host adapter, the micro-contract, SKILL.md, and the core contract.
• Proportional ceremony: short single-session answers skip checkpoints, draft files, and script audits; the file-backed bookend is reserved for long, multi-session, multi-agent, or durable-artifact work.

Absorbed reporting standards (all mappings in docs/REPORTING-STANDARDS.md)

• BLUF (AR 25-50), SBAR, Minto pyramid, US plain-language / ISO 24495-1, NN/g scanning research, WCAG 2.2 heading structure, Google SRE postmortems, IBCS SUCCESS / ISO 24896:2026.
• A two-round scientific-reporting literature survey: ASA p-value discipline, effect sizes with intervals, causal-language ladder, CONSORT/PRISMA accounting, uncertainty-visualization and error-bar semantics, benchmarking-crimes countermeasures.
• De-AI tone capability distilled from the MIT-licensed shuorenhua skill: a fidelity contract (tone edits never change facts), CN/EN signal-to-action table, misfire protection for technical vocabulary.

New surface area

• 3 display modules: ablation, benchmarking, natural-tone (5 → 8).
• 4 template assets: sbar-handoff, executive-onepager, rebuttal-response, release-card (8 → 12), each with an audit-clean finished example and receipts under examples/templates-2026082{6,8}/.
• 7 deterministic audit warnings: generic-heading, heading-level-skip, long-sentence (CJK-aware), deep-list-nesting, cjk-halfwidth-punctuation, ai-tone-boilerplate, plus the chart-first slide contract; all silent on known-good fixtures and showcase reports.
• Tightened slide surface contract from showcase feedback: 3+ data points must be charted with the takeaway encoded inside the visual.

Repository

• READMEs restructured after ARIS open-source conventions (numbered contents, dated What's New with PR links, collapsible release history, acknowledgements, star history).
• Evidence-first GitHub Pages landing page with the showcase deck PDF.
• reportctl --version, CITATION.cff, and the skill card are back in sync at 0.7.0.

Verification

• 309/309 unit tests (Python 3.9), check_test_env preflight clean, benchmark harness smoke, and dist reproducibility all green at 2ade309.
• Protocol and template content are unchanged since PR #5; this release adds only the version bump, the Evidence status rewrite, and the changelog close.

Full detail: CHANGELOG.md

v0.6.0 — Version flag, test-env preflight, live demo site

Highlights

• Live demo site — https://asimfish.github.io/super_agent_presentation/ : a landing page plus the real showcase deck, assembled from committed artifacts only (no analytics, no remote assets), deployed by the new pages.yml workflow.
• reportctl --version — the CLI now reports its version, and a regression test binds it to the version declared in CITATION.cff.
• Test-environment preflight — scripts/check_test_env.py fails when file content has been evicted by a sync client (macOS dataless files / .icloud placeholders) and warns when the clone sits in a cloud-sync scope. Added after evicted iCloud content was observed making the subprocess-heavy study tests hang and fail non-deterministically.
• Contribution surface — issue templates and a PR checklist that mirror the CONTRIBUTING evidence gates; citation metadata (CITATION.cff) enables GitHub's "Cite this repository" button.

Compatibility

No changes to protocols, routing, template content, or dist/ output. The frozen skill-tree manifest digest fixture was refreshed after the reportctl change, following the same procedure as the v0.5.0 release.

Verification

302/302 unit tests and presentation_benchmark.py smoke green on Python 3.9/3.12/3.14 (CI) and locally.

v0.5.0 — Receipt-backed showcase examples, bilingual README, catalog docs

Highlights

• Real, receipt-backed examples: examples/showcase-20260825/ vendors a complete end-to-end showcase — one finished sample for each of the 12 core modes and 4 research profiles, a 7-page assertion–evidence academic HTML deck with a Chrome-printed PDF and per-page renders, plus every route/checkpoint/bundle/strict-audit/semantic-oracle receipt and an honest first-failure log. Run verdict: ALL PASS (16/16 strict audit, 16/16 semantic oracle, 7/7 manual page review). All facts are synthetic fixtures; rendering verified on macOS Chrome 151 only.
• Bilingual README: English README.md and Chinese README_CN.md, both with visual previews of the rendered deck.
• Catalog docs: generated docs/CATALOG.md inventories modes, modules, profiles, surfaces, and template assets, cross-linked to bounded routes and finished examples; examples/README.md documents provenance and boundaries.
• Contributing guide: CONTRIBUTING.md covers the test, dist-reproducibility, real-render regression, and evidence-gate requirements.
• Print fix (landed in 9ee0d15): the academic-talk-html template no longer clips the leading glyph of long CJK headings when Chrome prints at 16:9; locked by filled-deck real-render regression tests with long CJK and mixed-script title fixtures.

No protocol, routing, or bundle-content changes beyond the print fix; dist output is unchanged from 0.4.0.

Verification

• 291/291 unit tests (Python 3.9), benchmark harness smoke, and dist reproducibility all green at 646d42c.
• The evidence stance is unchanged: the checked-in pilot remains insufficient_evidence; this release claims no measured quality, readability, or efficiency gain. See BENCHMARK.md.

v0.4.0 — Source-backed research reporting templates

Adds a composable research reporting layer without loading every template into agent context.

Highlights:

• research-idea mode plus RL, embodied-AI, world-model, and VLA profiles
• detailed experiment, paper-idea, HTML/PPT-style, and Quarto Reveal.js assets
• source ledger covering venue guidance, primary papers, benchmark repositories, and presentation workflows
• deterministic profile routing, exact template retrieval, link-only asset index, and bounded bundles
• 290 local tests, install journey, byte-identical distribution check, and real 16:9 HTML print rendering

Boundary: structural tests and the checked-in harness do not establish improved readability, factual correctness, or model compliance.

v0.3.2 — Controller-verified checkpoint receipts

Highlights

• Adds controller-verified checkpoint receipts for the ordered create -> bundle -> audit reporting lifecycle.
• Binds the final response to the exact captured report bytes, checkpoint intent, predeclared artifact paths, and owner-only file modes.
• Preserves local-image Markdown targets through agent execution, controller re-audit, stored records, and blind evaluation packets.
• Keeps legacy v1.0 receipts validation-readable without treating missing evidence as verified.
• Excludes private checkpoint artifacts and receipts from blind packets, aggregates, and release bundles.

Verification

• 279 local unit tests passed.
• GitHub Actions passed on Python 3.9, 3.12, and 3.14 for both main and the v0.3.2 tag.
• Evaluation harness smoke passed all 14 fixture expectations.
• Distribution rebuild was reproducible.
• Independent review found no remaining Blocker/P0, Major, or required Minor findings.

Effectiveness boundary

This release validates implementation behavior, artifact integrity, and deterministic evaluation plumbing. It does not make a public claim that the framework improves report quality in real-model use. No paid or real isolated Codex v1.1 smoke run was performed for this release.

Super Agent Presentation v0.3.1

Highlights

• Replaces full-tree receipt materialization with bounded iterative scandir traversal.
• Counts every inspected entry toward the 4,096-entry cap, prunes pycache subtrees, rejects symlinks and nonregular entries before ignore rules, and checks the 32 MiB byte cap before hashing.
• Preserves canonical ordering and the existing receipt digest for accepted Skill trees.
• Upgrades immutable GitHub Actions pins to official Node 24 releases: checkout v7.0.1 and setup-python v7.0.0.

Evidence and claim boundary

The actual 48-entry Skill tree retained manifest digest abaa134681e6180ef59cc70ecf35332a5fa9cd50dc0dfcea7d0a7860e4d45749. A fixed oversized-tree workload now rejects at entry 4,097 instead of scanning and accepting 20,001 entries; this demonstrates the resource bound, not a universal throughput speedup.

Both main and tag CI passed on Python 3.9, 3.12, and 3.14: 268 unit tests per matrix job, evaluation-harness smoke, route-bundle smoke, and byte-for-byte distribution reproducibility. Independent review found no release blockers.

The checked-in pilot remains insufficient evidence for quality, readability, or efficiency improvement. See CHANGELOG.md, PROJECT.md, SECURITY.md, and docs/PERFORMANCE.md for details.

Super Agent Presentation v0.3.0

Highlights

• Adds a standard-library-only preregistered study controller for immutable plans, paired generation records, blinding, rating freeze, aggregation, and fail-closed claim gates.
• Adds a typed Codex host adapter with inert planning, explicit execution authorization, fixed shell-free argv, executable/workspace/Skill receipts, bounded captures, and honest unsupported-control reporting.
• Separates caller and controller-stored generation schemas, adds portable generated-image ingest, and binds records, responses, transcripts, host receipts, assignment keys, blind packets, and ratings against drift.
• Adds public-design gates for held-out coverage, external per-unit isolation, multiple model revisions and repeats, context/compaction observations, final audit, display checks, visual precision/recall, and semantic-slot density per 1,000 output tokens.
• Removes the redundant automatic conclusions module from experiment reports; the default experiment bundle is now 11,597 characters while retaining calibrated conclusion guidance.

Evidence and claim boundary

The checked-in Codex pilot is a one-case, one-repeat integration and activation observation. It remains permanently marked insufficient_evidence and does not establish quality, readability, or efficiency improvement.

Release verification passed on Python 3.9, 3.12, and 3.14: 263 unit tests per matrix job, fixture smoke, route-bundle smoke, and byte-for-byte distribution reproducibility. Independent review found no release blockers.

See CHANGELOG.md, BENCHMARK.md, and docs/adr/ADR-006-preregistered-study-pipeline.md for the complete protocol and limitations.

Super Agent Presentation v0.2.0

Highlights

• Adds schema-v2 full-intent checkpoints and a same-checkpoint final audit for long-running tasks.
• Adds bounded rendered-text anchors in conservative plain Markdown prose, with conflict, drift, and diagnostic-leak protections.
• Unifies audit and benchmark image parsing behind one self-contained bounded scanner.
• Adds input, normalization, numeric-token, and resource guards tested on Python 3.9 and 3.14.
• Aligns host adapters, generated route packs, architecture, security, performance, and release documentation.

Verification

• 216/216 tests pass on Python 3.9 and Python 3.14.
• Smoke harness: 14/14 fixture expectations, 7/7 route expectations, and 5/5 post-activation route proxies.
• Distribution rebuild is byte-for-byte reproducible; clean isolated installation passes.

The benchmark is structural and harness-only. It does not claim measured improvement in real-host activation or human-rated report quality.

Super Agent Presentation v0.1.0

Initial structural release of the agent-native reporting framework.\n\n- 11 routed report modes and 5 on-demand display modules\n- persistent micro-contract with long-task checkpoints and final reload\n- structural audit plus optional validated JSON report IR\n- project installers for AGENTS, Codex, Claude, Cursor, and Copilot\n- deterministic activation, fixture, resource, and distribution checks\n\nRelease gates pass on Python 3.9, 3.12, and 3.14. The included evaluations are structural/development evidence and do not establish general model-effectiveness gains.