Chara's Toolkit v1.0.0

• Add original rust-engineering guidance for bounded crate changes and source-language ports, with conditional API, ownership/runtime, unsafe/FFI, parity, and supported-configuration references.
• Add a Rust checklist to the existing review specialist, preserving review, diagnosis, migration, and release ownership.
• Extend routing, overlap, workflow, behavioral probes, catalogs, and isolated-install coverage for the Rust route. Evaluation expectations are distinct from observed runs and comparative performance claims.
• Add original Kotlin, Python, C++, Swift, and Go implementation skills and self-contained language checklists for the existing reviewer. Framework guidance is conditional: Android/Compose/KMP, Qt/QML, and Apple/Metal retain their actual platform contracts.
• Expand routing, workflows, behavioral cases, catalogs, and isolated-install coverage to 28 installable skills. Preserve review, causal debugging, migration, security, performance, and release ownership across mixed-language projects.

Chara's Toolkit v0.10.0

• All shipped content is now original work. Rewrite every skill and reference that previously adapted third-party material, remove THIRD_PARTY_NOTICES.md, and add a validator guard that rejects reintroduced notices or upstream markers.
• Breaking: unlazy is replaced by completion-proof, a new completion-discipline method (done contract, coverage accounting, final-candidate pass). chara setup moves a toolkit-installed unlazy copy into the backup folder and leaves a user-owned skill of the same name untouched.
• Replace the vendored Oxlint runtime with chara-lint, 15 original rules written from a clean-room behavioral spec as dependency-free .mjs modules for Oxlint 1.89. Rule names changed (see assets/chara-lint/README.md). New options: allowEnvironmentChecks on no-typeof-checks; terms and ignoreForeignMembers on no-banned-identifier-terms. Open-dictionary checks now ignore finite-key records and homomorphic mapped types. A fixture suite runs every case through real Oxlint in CI.
• Agent-neutral setup. chara setup installs skills into the standard ~/.agents/skills folder and into every detected agent (Claude Code, Cursor, Gemini CLI, GitHub Copilot, OpenCode, Windsurf, Factory Droid, Kiro CLI, Roo Code, Goose); choose explicitly with --agents, and use --home for a different base folder. The Codex integration (Mission Control roles, AGENTS.md routing) installs only when Codex is targeted. Toolkit copies in the deprecated $CODEX_HOME/skills move to the backup folder so Codex does not list them twice; user-created skills stay. The updater remembers --home and --agents.
• Mission Control on every agent. The six roles are now neutral role cards inside delegate-with-mission-cards; scripts/build_mission_control.mjs generates native subagents for Codex, Claude Code, and Gemini CLI, and setup installs them where those agents are present. Other agents use the cards through a generic subagent or the main session. Read-only roles get a tool allowlist in Claude Code and Gemini CLI. Codex role files drop sandbox_mode, which Codex ignores in custom agent files, and state access in their instructions instead.
• Fix the documented doc_drift_scan.py command in documentation-synchronizer, which failed without --base or --files, and name every handoff target by its exact skill name.
• Remove time-sensitive content from skills, following Anthropic's skill-authoring guidance: the four dated ecosystem/platform snapshots become durable selection guides (package-selection.md for web, Flutter, Expo/React Native; platform-decision-matrix.md for mobile) that end with a checklist for verifying current versions, support, and advisories at decision time. Dated phrasing elsewhere in references is gone. The validator now rejects dated snapshot files in skills instead of requiring them, which also removes the review-deadline failure mode that blocked the 0.9.6 release job.
• Add 27 behavioral decision-boundary probes, three for each of the nine skills that had none, so every evaluated skill is checked for outcome, not only for routing.
• Every skill description now states when not to use the skill, following the published Anthropic and OpenAI skill-authoring guidance; reference files over 100 lines start with a contents list.
• Scope typescript-quality-enforcer to JavaScript/TypeScript only, with routing cases that keep it away from Python and Dart lint work.
• Rewrite product-design-director, screenshot-to-interface (new measure_tokens.py helper for contrast, color-token snapping, and spacing scale), and security-review (new finding calibration).
• Make content-provenance-hygiene backend-agnostic: a local inspection-toolchain reference and a read-only inspect_text_invisibles.py scanner replace the single-service protocol.
• Rewrite the flaky-test, characterization-testing, dependency-risk, database-migration, GitHub Actions, Core Web Vitals, and TypeScript strictness references.

Chara's Toolkit v0.9.6

• Upgrade PR review with evidence-checked findings, risk-adjusted depth, optional independent verification, and a three-round convergence stop (#22).

• Add opt-in review policy guidance, repository REVIEW.md, PR template, and regression-focused routing/workflow evaluations.

• Refresh Next.js, Expo/React Native, Flutter, and mobile architecture references against current releases/security advisories; make ordinary CI use today's UTC date for freshness checks, matching release verification.

• Add opt-in project verification with live readiness checks and explicit execution acknowledgment.

• Add signed managed-install approval, cross-platform destructive-operation guards, and offline behavioral/outcome evaluation helpers.

• Add recurring-error auditing and measured GitHub Actions guidance while keeping claims bounded to available evidence.

Chara's Toolkit 0.9.5

• feat: rebrand as Chara's Toolkit (#21) (113430a3d30f).

Codex Toolkit 0.9.4

• feat: universalize coordination and Mission Control tooling (278d15739005).

Codex Toolkit 0.9.3

• feat: make Mission Control model-agnostic (321088eaac83).

Codex Toolkit 0.9.2

• docs: simplify README for new users (9be2460c7b52).

Codex Toolkit 0.9.1

• docs: add canonical README hero (b32581bb2a45).

Codex Toolkit 0.9.0

• Add security-review as a focused security code-review specialist with trust-boundary tracing, filtered exploit findings, and regression-oriented fixes.
• Compress all production SKILL.md roots and routing descriptions around progressive disclosure, preserving conditional specialist detail in references instead of default context.
• Fold flaky-test diagnosis, characterization testing, dependency risk, production database migration safety, GitHub Actions hardening, TypeScript strictness migration, and Core Web Vitals guidance into their existing owning skills.
• Modernize skill validation with concise-description and root-size review signals while preserving Agent Skills hard ceilings and routing/evidence checks.
• Extend routing, overlap, workflow, responsibility, installer, and release metadata coverage to 22 installable skills and 21 production routes.
• Preserve MIT attribution and source mapping for the SkillMedev/skills concepts adapted in this release.

Codex Toolkit 0.8.8

• docs: define evaluation evidence agent contracts (e700ed2c7878).