Skip to content

Releases: skymanbp/cc-tree

v0.2.0 — self-contained + roadmap features

29 May 15:39

Choose a tag to compare

Closes the v0.1 doc/impl drift and lands the EVALUATION roadmap. cc-tree is now CLI-installable and self-contained.

Fixed (doc/impl drift)

  • Preset schema validator now enforces what the docs promised. New zero-dependency tools/_frontmatter.py + tools/validate_plugin.py enforce all 8 rules (node_schema=12, score_dims=5 w/ key/name/desc, verdict_enum 4 roles, convergence_metric ∈ roles, known root_kind, output_artifacts.primary, name=basename). tools/test_validate.py proves it (4 presets pass + 9 negative cases rejected); both run in CI.
  • §3.X external-check is now preset-aware (attack→errata/critiques, code-audit→CVE/advisory, brainstorm/design→prior art + tooling).
  • Resolved the convergence_metric alias contradiction in docs/presets.md.
  • README install steps corrected to the real plugin CLI; attack/code-audit verdict vocabulary fixed (INCOMPLETE_FORBIDDEN, not DEAD-END).

Added (roadmap)

  • --field <name|path> — domain-aware reviewer weighting via field-profiles/ (feeds §3.C/§3.D/§3.I/§3.J; non-blocking if absent).
  • Cross-preset chaining/cc-tree:tree-chain + docs/chaining.md + universal --seed-from (alias --from-prior).
  • Mandatory sub-agent dispatch at fan-out ≥ 5 (docs/ENGINE.md §8.1).
  • Bilingual engine docsdocs/ENGINE.zh.md + docs/framings.zh.md (English canonical).
  • Concept diagram in README (ASCII radial tree + Mermaid engine loop).
  • examples/attack/ illustrative run.

Full detail: CHANGELOG.md.

v0.1.0 — Universal radial-tree exploration engine

25 May 17:15

Choose a tag to compare

Initial release of cc-tree — a Claude Code plugin that extracts the universal radial-tree exploration pattern from skymanbp/sci-paper's brainstorm + paper-attack-tree skills, strips the paper-specific anchors, and generalizes it via swappable preset files.

What ships

  • 1 skill /cc-tree:tree — universal engine
  • 4 presets: brainstorm / attack / design / code-audit
  • 4 ergonomic slash-commands: /cc-tree:brainstorm, :attack, :design, :code-audit
  • Full engine spec at docs/ENGINE.md (~800 lines covering §0 forbidden patterns through §11 compliance checklist)
  • 12 framings detail at docs/framings.md (§3.A–§3.L with per-preset examples)
  • Preset authoring guide at docs/presets.md
  • Validator + CI gating manifest alignment, frontmatter, and tools syntax

Key behavioral commitments

All preset-agnostic and unweakening:

  • Hard ban on incomplete leaves (§F8): no defer / future-work / TODO / NEEDS-MORE-INFO allowed — every leaf is fully derived before counting toward final width.
  • All 12 framings fire per node per round (§F4 + --min-frameworks 12): each of First-principles / Inversion / Cross-disciplinary / Adversarial / Constraint-variation / Scale-extrapolation / Substitution / Office-hours-6Q / Contrarian / Failure-driven / High-risk-asymmetric / Meta-self-audit produces ≥ 1 child.
  • Substantive convergence is the intended termination (§6 six conditions all hold), not resource caps. Caps (--width / --depth / --rounds) default to ∞ and exist as escape valves.
  • Incremental write per node (§7.1) — crash-safe; resume by re-invoking with the same --out <dir>.

Provenance

Refactored from sci-paper's brainstorm and paper-attack-tree skills. sci-paper continues to ship its domain-specific versions independently; cc-tree is the domain-agnostic generalization.

See CHANGELOG.md and EVALUATION.md for full design rationale.