Consolidated Intelligence skills into seven shipped workflows

[dzykovic]

Summary

Reduced the shipped Intelligence catalog from twelve skills to seven. Context updating owns artifact authoring; session learning includes workflow extraction; review includes compaction; adapter management operates existing targets. Adapter implementation now lives in the repository-only dev-build-adapter skill, with the public CLI and adapter guide still available to consumers.

The version-management skill is named intelligence-upgrade to distinguish it from intelligence-update-context; its procedure and CLI commands remain unchanged. Added the skill-name migration map and decision 0008, updated agent bindings and onboarding, and prepared 0.15.0 with regenerated project/example locks and tool output. Skill content including references is 22.7% smaller (52,905 to 40,898 bytes).

Type of change

• CLI lifecycle or package behavior
• Sync engine or adapter
• Rule, agent, or skill content
• Documentation
• Refactor with no behavior change

Public CLI check

• Uses only the established groups: init, sync, update, upgrade, package, adapter, status, registry
• Keeps preview/apply behavior explicit for planned writes
• Covers automatic project alignment and CI refusal when project schema/content can change
• Keeps fresh-clone package restoration behind intelligence sync

The executable command model is unchanged. Existing lifecycle coverage runs alongside the new catalog assertions.

Verification

• bash cli/tests/verify.sh reports verify ok and skipped nothing this change needed
• intelligence status --check succeeds in this project after alignment to 0.15.0
• A second intelligence sync leaves all 84 generated/shared output files unchanged
• User-facing docs and examples use the current public command groups

The lifecycle regression checks the seven package and generated skills, conditional references, retired-command cleanup, local skill discovery, and repeatable rendering. bash npm/build.sh plus npm pack --dry-run --json confirms seven shipped skills and no repository-only adapter workflow. Source resource links and agent bindings resolve; the outgoing secret scan is clean. No adapter implementation changed, so real-tool adapter conformance testing does not apply.

The rename regression also removes obsolete intelligence-update output. The generic Codex skill validator rejects the repository's pre-existing agent and argument-hint extensions; repository-aware frontmatter and binding checks, generated output, and lifecycle tests cover those fields. The existing Windows filesystem probe skips its escaping-symlink fixture; Linux and macOS CI provide the platform checks.

Versioned artifacts

• The PR records its change under one concrete CHANGELOG.md version; there is no [Unreleased] section or duplicated release date
• The current engine/VERSION release state was checked: v0.14.0 is published, so this change prepares 0.15.0
• Every example's schema_version and exact sync pin changed with the engine version; committed locks were regenerated through the CLI
• No release is published by this PR

Notes for reviewers

The removed skill names have no registered aliases. The migration map covers invocations, bindings, and deliberate downstream overrides. Guided adapter implementation is intentionally repository-only; no development package is added.

Preservation was reviewed against the previous authoring, learning, compaction, onboarding, cleanup, and verification responsibilities. Automated checks cover distribution and lifecycle behavior; no model-based behavior evaluation was available.

Backout before release is one PR revert. For an already-updated consuming project, restore its prior source/manifest/lock snapshot and matching CLI version, then sync; generated output is reproducible from those sources.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.