Humanities Superpowers v2.1.0 — Dual-Core

Released 2026-09-15.

Humanities Superpowers v2.1.0 promotes the verified Dual-Core work into a release. It is not a new research feature: the Humanities/Fricturn research core keeps its 13 core skills, its router semantics, its verification states, its gate statuses, its validity rules, and its human authorization boundary.

What v2.1.0 contains

One router, two cores, and an explicit boundary between them.

                  Router
                     |
        -----------------------------
        |             |             |
     RESEARCH       CODE        HYBRID
        |             |             |
   Humanities   Engineering    Both cores
      Core          Core

Added

• Dual-Core architecture — a research core and an engineering core reached through one router, with an explicit domain boundary.
• Engineering core — ten selected obra/superpowers 6.3.0 skills, vendored byte-identically with per-skill hashes, import commit, license, and a documented reason for every excluded skill.
• RESEARCH / CODE / HYBRID routing — semantic classification by the artifact that must be correct, not by keywords.
• QUICK / STANDARD / STRICT risk routing — applied to code and hybrid work, derived from reversibility, blast radius, privilege boundary, persistence, contract surface, environment, failure cost, and uncertainty.
• Autonomous Worker state contract — WORKING, WAITING_INPUT, WAITING_PRIVILEGE, BLOCKED, ERROR, DONE, where a live process is not WORKING and a terminated process is not DONE.
• WORK_PACKAGE / EVIDENCE_PACKAGE — handoff records between role-based PLANNER, IMPLEMENTER, REVIEWER, and ESCALATION_REVIEWER assignments, with model names confined to a replaceable operating profile.
• Coding evaluation harness — scripts/hsp_eval_coding.py and tests/eval/coding-cases.json for the separate CONTROL_CODING versus ENGINEERING_CORE comparison.

Preserved

• The 13 core research skills are unchanged from the audited tip (2a452b6) apart from the single release-metadata version: line; the skill bodies, gates, and validity rules are identical.
• Gate status, current validity, and progression authorization stay separate properties, and invalidation remains bounded by dependency references.
• Epistemic return, citation verification, productive refusal, terminology control, and human judgment authority are unchanged.
• Installation safety: an existing AGENTS.md or CLAUDE.md is never overwritten, and skills/skills/ is never created.
• vendor/obra-superpowers/** is byte-identical to upstream commit b36e0829c6d0140e93cfef2ca599b1b07d4a7797.

Limitations

• Coding effectiveness: NOT_MEASURED. No CONTROL_CODING versus ENGINEERING_CORE run has been recorded, so this release claims no coding-quality improvement, no defect reduction, and no token saving.
• Research effectiveness: NOT_MEASURED. No CONTROL versus HSP run has been recorded, so this release claims no research-quality improvement.
• Runtime notification: not implemented. The worker state contract is a contract. No Telegram, Hermes, or other notification runtime ships with this release.
• Cursor runtime loading: NOT_MEASURED. The Cursor project layout is checked statically, but the rule has not been loaded in a target Cursor environment.
• Passing repository, installation, and conformance checks demonstrate internal consistency. They are not evidence of benefit.

Version scope

v2.1.0 is the project release version. It is not the Fricturn protocol version: the v2 specification documents remain at Specification Version: 2.0.0, schema identifiers are unchanged, and schema_version values in protocol records are unchanged.

Provenance

Field	Value
Upstream repository	https://github.com/obra/superpowers
Upstream version	6.3.0
Upstream commit	b36e0829c6d0140e93cfef2ca599b1b07d4a7797
Upstream license	MIT, Copyright (c) 2025 Jesse Vincent
Imported skills	10 of 14
Record	vendor/obra-superpowers/PROVENANCE.json

The vendored files are unmodified. scripts/run_dual_core_tests.py recomputes the recorded hashes and fails if any vendored byte changes.

Validation

The release candidate passed repository validation, integrated tests, v2 conformance tests, dual-core routing and provenance tests, public-release validation, installation smoke tests, both evaluation self-tests, and a strict documentation build. Manifest hashes were regenerated from the release tree.

Codex and Claude Code clean installs were verified: the documented commands were executed in isolated projects and reproduced 14 Humanities skills, 10 vendored engineering skills, the router, templates/, schemas/, the dual-core specification, and the provenance record, without overwriting an existing AGENTS.md or CLAUDE.md and without creating skills/skills/.

The vendor tree, manifest, and the v2.1.0 tag were re-verified on main after the fast-forward merge, and the same validation suite passed again there.

Humanities Superpowers 2.0 — Fricturn

Humanities Superpowers v2.0.0 — Fricturn

Released 2026-08-21.

Fricturn extends the v1 scholarly judgment scaffold without changing its 13 core skills, one router, verification-state vocabulary, gate statuses, or routing decisions.

Added

• Interpretive friction as a traceable research event.
• Claim-specific evidential standing distinct from source verification.
• Append-only judgment and interpretation histories.
• Epistemic return for knowledge-driven reopening of earlier objects.
• Bounded gate invalidation through explicit dependency references.
• Productive refusal records that preserve checked range and next verification.
• Twelve deterministic v2 conformance cases, explicit negative boundary checks, and four synthetic examples.

Compatibility

v1 research objects and sessions remain valid because the extended base fields are optional. Dedicated v2 schemas enforce stronger object-level rules. Gate status remains PASS, CONDITIONAL PASS, or FAIL; validity is a separate property.

Validation

The release passed repository validation, integrated tests, v2 conformance tests, public-release validation, JSON/YAML parsing, link checks, strict documentation build, and manifest regeneration.

Humanities Superpowers v1.0.1

Humanities Superpowers v1.0.1

This patch improves public-release reliability and documentation clarity.

Improved

• Clarified project-local installation for Claude Code and OpenAI Codex
• Marked Cursor installation as guidance-only and not independently verified
• Standardized terminology as 13 core research skills plus 1 Level 3 router
• Added cross-platform validation for UTF-8, LF/CRLF, and path-case conflicts
• Expanded GitHub Actions validation to run all three validators and MkDocs strict build
• Improved manifest reproducibility and public-release checks
• Added safer first-run and read-only pilot guidance

Compatibility

Tested:

• Claude Code
• OpenAI Codex

Installation guidance is provided for Cursor, but Cursor loading has not yet been independently verified.

Methodology

No research skill contracts, quality-gate rules, or methodological behavior were changed in this patch.

Humanities Superpowers v1.0.0

Humanities Superpowers v1.0.0

Structured research skills and quality gates for humanities scholars using AI agents.

Not an AI paper writer. A scholarly judgment scaffold.

13 core skills

• Formulating research questions with iterative revision
• Medium-specific close reading (text, image, film, interface, performance)
• Concept-lineage tracing (distinguished from Foucauldian genealogy)
• Literature dialogue, argument planning, stress testing
• Citation auditing, terminology control, peer-review response
• Submission verification with PASS/CONDITIONAL PASS/FAIL gates

Key features

• Quality gates that can return FAIL — a framework that refuses to hide unresolved risk
• 8 verification states (verified-primary through disputed)
• Anti-fabrication rules embedded in every skill
• Hermeneutic part-whole iteration protocol
• Researcher steelman confirmation before adversarial testing
• Argument-type-specific objection prioritization
• Intentional example failure to demonstrate gate integrity
• White paper with verified citation audit

Installation

git clone https://github.com/icerain-cmd/humanities-superpowers.git
python3 scripts/validate_repository.py

See INSTALLATION.md for Claude Code, Codex, and Cursor setup.

License

MIT — inspired by Jesse Vincent's obra/superpowers, independently authored.