feat(re-anchor): add script-the-deterministic-work corrector (0.3.0)

.claude-plugin/marketplace.json

@@ -334,7 +334,7 @@
       "name": "re-anchor",
       "source": "./plugins/re-anchor",
       "category": "quality",
-      "tags": ["re-anchor", "discipline", "corrector", "research", "standards", "duplication", "incumbency", "terseness", "upstream", "conformance", "selection", "communication", "audit", "skill"]
+      "tags": ["re-anchor", "discipline", "corrector", "research", "standards", "duplication", "incumbency", "terseness", "upstream", "conformance", "selection", "communication", "determinism", "audit", "skill"]
     },
     {
       "name": "education",

README.md

@@ -95,7 +95,7 @@ user opts in with `/plugin enable`; an existing install is never flipped by cata
 - [`mcp-tools`](plugins/mcp-tools) — Audits MCP server tool definitions against MCP-specification and Anthropic tool-design criteria and reports a per-tool PASS/WARN/FAIL scorecard covering description, parameters, naming, and annotations. Language-agnostic — Python (FastMCP), TypeScript, and .NET.
 - [`review`](plugins/review) — Code-review toolkit: six read-only reviewer agents (code, security, architecture, doc drift, build/test/lint, CI-log audit) plus two orchestration skills — a single-lens quality gate and a multi-surface review fan-out with severity-ranked, deduplicated findings.
 - [`codebase-health`](plugins/codebase-health) — Repo-wide drift audit between docs, config, code, and architecture: verifies every factual claim against reality via parallel subagent fan-out, severity-rates findings, and fixes or presents for review. Audit dimensions are configurable through a tracked .claude/codebase-health.md config file written by the setup skill.
-- [`re-anchor`](plugins/re-anchor) — Discipline correctors that re-anchor a standing rule mid-session, then audit both the work in flight and the pre-existing state and choices it trusts, and correct what has drifted: do-your-research (research and no-assumptions discipline; sibling do-your-research-deep escalates to a subagent-fanned verification pass over every load-bearing claim), follow-our-standards (alignment to the consuming org's engineering conventions), point-dont-copy (pointer-over-copy discipline — no copied content, internal-name coupling, or closed capability lists), reason-dont-recite (interrogate inherited content — precedent is evidence of what is, never self-justifying authority), tighten-your-output (terseness discipline — fewer words or lines with no loss of meaning or correctness), recheck-against-upstream (existing state is not evidence of its own correctness — audit config, code, and infra against current official upstream docs; sibling recheck-against-upstream-deep fans subagents doc-by-doc over a whole subsystem), pick-for-the-problem (tool, library, framework, and approach selection fitted to the problem, not reached for out of habit, availability, incumbency, or preconception), and mind-your-maxims (cooperative-communication discipline per Grice plus the AI-augmented transparency maxim). Firing one is a re-anchor, not an accusation; the audit may return clean.
+- [`re-anchor`](plugins/re-anchor) — Discipline correctors that re-anchor a standing rule mid-session, then audit both the work in flight and the pre-existing state and choices it trusts, and correct what has drifted: do-your-research (research and no-assumptions discipline; sibling do-your-research-deep escalates to a subagent-fanned verification pass over every load-bearing claim), follow-our-standards (alignment to the consuming org's engineering conventions), point-dont-copy (pointer-over-copy discipline — no copied content, internal-name coupling, or closed capability lists), reason-dont-recite (interrogate inherited content — precedent is evidence of what is, never self-justifying authority), tighten-your-output (terseness discipline — fewer words or lines with no loss of meaning or correctness), recheck-against-upstream (existing state is not evidence of its own correctness — audit config, code, and infra against current official upstream docs; sibling recheck-against-upstream-deep fans subagents doc-by-doc over a whole subsystem), pick-for-the-problem (tool, library, framework, and approach selection fitted to the problem, not reached for out of habit, availability, incumbency, or preconception), mind-your-maxims (cooperative-communication discipline per Grice plus the AI-augmented transparency maxim), and script-the-deterministic-work (offload deterministic sub-work — counts, diffs, sorts, transforms, and scaffolds — to a script that runs, reserving model output for judgment over its real output). Firing one is a re-anchor, not an accusation; the audit may return clean.
 
 ### Maintenance
 

plugins/re-anchor/.claude-plugin/plugin.json

@@ -1,12 +1,12 @@
 {
   "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
   "name": "re-anchor",
-  "version": "0.2.0",
-  "description": "Discipline correctors that re-anchor a standing rule mid-session, then audit both the work in flight and the pre-existing state and choices it trusts, and correct what has drifted: do-your-research (research and no-assumptions discipline; sibling do-your-research-deep escalates to a subagent-fanned verification pass over every load-bearing claim), follow-our-standards (alignment to the consuming org's engineering conventions), point-dont-copy (pointer-over-copy discipline — no copied content, internal-name coupling, or closed capability lists), reason-dont-recite (interrogate inherited content — precedent is evidence of what is, never self-justifying authority), tighten-your-output (terseness discipline — fewer words or lines with no loss of meaning or correctness), recheck-against-upstream (existing state is not evidence of its own correctness — audit config, code, and infra against current official upstream docs; sibling recheck-against-upstream-deep fans subagents doc-by-doc over a whole subsystem), pick-for-the-problem (tool, library, framework, and approach selection fitted to the problem, not reached for out of habit, availability, incumbency, or preconception), and mind-your-maxims (cooperative-communication discipline per Grice plus the AI-augmented transparency maxim). Firing one is a re-anchor, not an accusation; the audit may return clean.",
+  "version": "0.3.0",
+  "description": "Discipline correctors that re-anchor a standing rule mid-session, then audit both the work in flight and the pre-existing state and choices it trusts, and correct what has drifted: do-your-research (research and no-assumptions discipline; sibling do-your-research-deep escalates to a subagent-fanned verification pass over every load-bearing claim), follow-our-standards (alignment to the consuming org's engineering conventions), point-dont-copy (pointer-over-copy discipline — no copied content, internal-name coupling, or closed capability lists), reason-dont-recite (interrogate inherited content — precedent is evidence of what is, never self-justifying authority), tighten-your-output (terseness discipline — fewer words or lines with no loss of meaning or correctness), recheck-against-upstream (existing state is not evidence of its own correctness — audit config, code, and infra against current official upstream docs; sibling recheck-against-upstream-deep fans subagents doc-by-doc over a whole subsystem), pick-for-the-problem (tool, library, framework, and approach selection fitted to the problem, not reached for out of habit, availability, incumbency, or preconception), mind-your-maxims (cooperative-communication discipline per Grice plus the AI-augmented transparency maxim), and script-the-deterministic-work (offload deterministic sub-work — counts, diffs, sorts, transforms, and scaffolds — to a script that runs, reserving model output for judgment over its real output). Firing one is a re-anchor, not an accusation; the audit may return clean.",
   "author": {
     "name": "Melodic Software",
     "email": "info@melodicsoftware.com"
   },
   "license": "MIT",
-  "keywords": ["re-anchor", "discipline", "corrector", "research", "standards", "duplication", "incumbency", "terseness", "upstream", "conformance", "selection", "dependencies", "communication", "skill"]
+  "keywords": ["re-anchor", "discipline", "corrector", "research", "standards", "duplication", "incumbency", "terseness", "upstream", "conformance", "selection", "dependencies", "communication", "determinism", "scripting", "skill"]
 }

plugins/re-anchor/CHANGELOG.md

@@ -3,6 +3,25 @@
 All notable changes to the `re-anchor` plugin are documented here. Format follows
 [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); this plugin uses semantic versioning.
 
+## [0.3.0]
+
+### Added
+
+- `/re-anchor:script-the-deterministic-work` — offload-the-deterministic
+  discipline: purely deterministic sub-work (counting, diffing, sorting,
+  transforming, matching, sweeping, arithmetic) gets a script that runs and
+  returns real output, and the model reasons only afterward over that output.
+  The tier boundary — deterministic (script it), detect-then-judge (script
+  the detect half; the verdict stays judgement), reasoning-only (never
+  script) — re-anchors the consuming org's enforceability-tiers convention;
+  the in-task "script it now" application has no standards doc yet, so the
+  skill flags that gap rather than inventing a rubric. Runs in both
+  directions: analysis reasons over a script's output; generation emits a
+  deterministic scaffold (PR body, issue, report, config boilerplate) from a
+  script or native template so model output is reserved for the judgment
+  slots. Distinct from a standing-automation capability: recurring checks
+  route to a hook, this corrector owns the one-off, session-time script.
+
 ## [0.2.0]
 
 ### Added

plugins/re-anchor/README.md

@@ -27,6 +27,7 @@ first-class use, and the audit may honestly return clean.
 | `/re-anchor:recheck-against-upstream-deep` | The fan-out tier of recheck-against-upstream — subagents compare a whole subsystem against upstream, doc-by-doc |
 | `/re-anchor:pick-for-the-problem` | Selection fitted to the problem — not reached for out of habit, availability, incumbency, or preconception |
 | `/re-anchor:mind-your-maxims` | Cooperative communication — Grice's maxims plus the AI-augmented transparency maxim |
+| `/re-anchor:script-the-deterministic-work` | Script deterministic sub-work — run it, then reason over the output |
 
 The shared method — re-anchor, audit the work in flight, correct forward,
 report — lives once at plugin scope in
@@ -202,6 +203,28 @@ as posture-setting anytime and as an audit once output exists.
 /re-anchor:mind-your-maxims        # re-anchor + audit + correct
 ```
 
+### script-the-deterministic-work
+
+Re-anchors the discipline of offloading deterministic sub-work to a script:
+when a sub-task's answer follows mechanically from its input (counting,
+diffing, sorting, transforming, matching, sweeping, arithmetic), write and
+run a script, read its real output, and reason only afterward over that
+output. The tier boundary re-anchors the consuming org's enforceability-tiers
+convention — deterministic work gets scripted, detect-then-judge gets only
+its detect half scripted while the verdict stays judgement, and
+reasoning-only is never scripted. The in-task "script it now" application has
+no standards doc yet, so the skill flags that gap. The discipline runs in
+both directions: analysis reasons over a script's output, and generation
+emits a deterministic scaffold (a PR body, an issue, a report, config
+boilerplate) from a script or a native template so model output is reserved
+for the judgment slots. Distinct from a standing-automation capability:
+recurring checks belong in a hook, this corrector owns the one-off,
+session-time script.
+
+```shell
+/re-anchor:script-the-deterministic-work   # re-anchor + audit + correct
+```
+
 ## Consumer conventions
 
 The correctors adapt to the consuming repo rather than imposing a source