From cef0fefd2a4893a330a5216b4b8e5d68a19c2ccc Mon Sep 17 00:00:00 2001 From: Sascha Corti Date: Wed, 3 Jun 2026 11:25:06 +0200 Subject: [PATCH 1/2] feat: add experimental Caveman response style skill Integrates an ultra-compressed response style that reduces output token count while preserving technical accuracy. This skill includes configurable intensity levels and auto-clarity safety rules. The skill is added to the experimental and hve-core-all collections and plugins. feat(caveman): Explicitly declare maturity and add attribution Sets the `caveman` skill's maturity to `experimental` in the collection manifest, aligning with its status. Additionally, updates the plugin README with attribution for the skill's origin. feat(caveman): Define experimental maturity and refine attribution Sets the 'caveman' skill's maturity to 'experimental' in the collection manifest. The explicit attribution in the plugin README is removed, with the author's surname 'Brussee' added to the spell-check dictionary instead for consistency. Also corrects minor table formatting in skill documentation. feat(caveman): Explicitly declare maturity and add attribution Sets the `caveman` skill's maturity to `experimental` in the collection manifest, aligning with its status. Additionally, updates the plugin README with attribution for the skill's origin. feat(caveman): Define experimental maturity and refine attribution Sets the 'caveman' skill's maturity to 'experimental' in the collection manifest. The explicit attribution in the plugin README is removed, with the author's surname 'Brussee' added to the spell-check dictionary instead for consistency. Also corrects minor table formatting in skill documentation. docs(skills): normalize caveman auto-clarity list, clarify wenyan combo Normalize the Auto-Clarity Boundaries bullet list to uniformly complete sentences ending with periods, addressing the markdown punctuation consistency rule. Replace the abstract wenyan-combination wording with an explicit `/caveman wenyan ultra` example so readers can see what a valid level combination looks like. Resolves the Major and Minor findings from the implementation review at .copilot-tracking/reviews/2026-06-05/caveman-skill-enhancements-plan-review.md. revert(scope): restore adr-author SKILL.md and activation baseline to main Commits b37cac9d and 1595c7d4 (caveman feature work) incidentally ran \`npm run format:tables\` over the repo and tightened markdown table padding in .github/skills/project-planning/adr-author/SKILL.md. That file is loaded by the @adr-creation GovernEntry and AdoptTemplate activation scenarios, which pin exact byte counts and content hashes in scripts/agents/activation-harness/baseline.json. The drag-along change broke 4 Pester tests in CI and forced an out-of-scope baseline recapture (commit 470c4f4f). Restore both files to their state on origin/main so this branch contains only caveman-related changes. The adr-author SKILL.md formatting drift is unrelated to the caveman skill and should land via its own PR if desired. chore(baseline): Recalibrate activation harness for adr-author skill Updates the pinned byte counts and content hashes in `baseline.json` for the `GovernEntry` and `AdoptTemplate` scenarios. This recalibration is necessary due to recent modifications to `.github/skills/project-planning/adr-author/SKILL.md`, which altered its size and content. chore(plugins): regenerate plugin outputs and re-baseline activation harness Postprocess (markdown-table-formatter) normalizes tables in caveman and adr-author SKILL.md, plus the experimental collection and plugin README. Re-baselines activation harness against canonical adr-author SKILL.md (13531 bytes). Update .github/skills/experimental/caveman/SKILL.md Co-authored-by: Chris Montazer <17170709+rezatnoMsirhC@users.noreply.github.com> --- .cspell/general-technical.txt | 1 + .github/skills/experimental/caveman/SKILL.md | 105 ++++++++++++++++++ collections/experimental.collection.md | 1 + collections/experimental.collection.yml | 3 + collections/hve-core-all.collection.yml | 3 + plugins/experimental/README.md | 1 + .../experimental/skills/experimental/caveman | 1 + .../hve-core-all/skills/experimental/caveman | 1 + 8 files changed, 116 insertions(+) create mode 100644 .github/skills/experimental/caveman/SKILL.md create mode 120000 plugins/experimental/skills/experimental/caveman create mode 120000 plugins/hve-core-all/skills/experimental/caveman diff --git a/.cspell/general-technical.txt b/.cspell/general-technical.txt index 06a0496d2..e11b8def4 100644 --- a/.cspell/general-technical.txt +++ b/.cspell/general-technical.txt @@ -1621,3 +1621,4 @@ transitioner urlsafe wberry writeback +Brussee diff --git a/.github/skills/experimental/caveman/SKILL.md b/.github/skills/experimental/caveman/SKILL.md new file mode 100644 index 000000000..80bb2743a --- /dev/null +++ b/.github/skills/experimental/caveman/SKILL.md @@ -0,0 +1,105 @@ +--- +name: caveman +description: 'Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules - Brought to you by microsoft/hve-core' +license: MIT +disable-model-invocation: true +metadata: + authors: "microsoft/hve-core" + spec_version: "1.0" + last_updated: "2026-06-05" + content_based_on: "https://github.com/JuliusBrussee/caveman" +--- + +# Caveman Skill + +## Overview + +Caveman is an opt-in response style that reduces output verbosity while keeping technical content fully intact. The agent drops articles, filler words, hedging, and pleasantries; keeps fragments where they remain unambiguous; and writes code, error messages, identifiers, and command-line arguments verbatim. Use it when the user explicitly requests a terser response. + +The concept originates from the upstream Caveman project by Julius Brussee (MIT licensed; see Attribution). This skill is an original specification of that behavior and ships no upstream files. + +## How the Mode Persists + +Caveman has no out-of-band state store, daemon, or hook. Persistence relies entirely on the conversation transcript: + +* The activation message (`/caveman ultra`, "use caveman", and similar) stays visible in chat history. +* On each turn, read the most recent activation, exit, or level-switch directive in the transcript and apply the corresponding tone. The latest matching directive wins. +* The skill file is loaded on demand. Once the rules are in context, keep applying them without reloading. If context is trimmed and the rules drop out, reload `caveman/SKILL.md` the next time an active directive appears. +* If the transcript is cleared, the conversation ends, or the activation message falls out of scope, the mode is off by default. The user re-invokes to turn it back on. + +State lives in chat, not in a file. If the activation is not visible in the transcript, the mode is not active. + +## When to Use + +Activate Caveman when the user asks for it directly: + +* "use caveman", "caveman mode", "talk caveman" +* `/caveman` or `/caveman ` where `` is one of `lite`, `full`, `ultra`, `wenyan` + +Do not activate on generic brevity requests such as "be brief", "less tokens", "terser output", or "save tokens". Those are one-shot asks for the current reply, not requests to flip a persistent mode. + +Stop Caveman when the user says "stop caveman", "normal mode", "verbose again", or `/caveman off`. + +## Intensity Levels + +| Level | Behavior | +|------------------|------------------------------------------------------------------| +| `lite` | Drop filler and hedging. Keep articles and full sentences. | +| `full` (default) | Drop articles. Sentence fragments allowed. Short synonyms. | +| `ultra` | Telegraphic. One-word answers when sufficient. Arrows for flow. | +| `wenyan` | Classical Chinese (文言) register layered on `full` compression. | + +If the user requests `/caveman` without a level, default to `full`. `/caveman wenyan` applies the wenyan register at `full` compression. Combine with another level for stronger compression, e.g. `/caveman wenyan ultra`. + +## Compression Rules + +Always drop: + +* Articles such as a, an, the +* Filler words such as just, really, basically, simply, actually +* Pleasantries such as "happy to help", "great question", "of course" +* Hedging phrases such as "you might want to", "perhaps consider", "it could be" + +Always keep, exact and unmodified: + +* Code blocks +* Function, class, variable, file, and command names +* Error messages and stack traces +* CLI flags and configuration values +* URLs and file paths + +Pattern: `[thing] [action] [reason]. [next step].` + +## Auto-Clarity Boundaries + +Switch off Caveman automatically — without being asked — when any of the following apply, then resume after the section ends: + +* Security warnings or vulnerability disclosures are being communicated. +* Confirmations are required for destructive or irreversible actions such as delete, drop, force push, or rm -rf. +* Multi-step sequences are involved where dropping conjunctions would create order ambiguity. +* Tool output is being quoted, such as linter warnings, test failures, terminal errors, CI logs, and stack traces. Quote verbatim — these can carry safety-relevant detail (for example, a linter flagging a hardcoded secret) that compression would erase. +* The user appears confused or asks for clarification — drop to normal until clarity is restored, then resume the previously selected level. +* Compression would make a technical instruction ambiguous. + +Code, commits, pull request bodies, and release notes are always written in normal style regardless of mode. + +## Examples + +Normal: "I'd be happy to help! The bug is most likely in your authentication middleware where the token expiry check uses a strict less-than comparison." + +Caveman (full): "Bug in auth middleware. Token expiry check uses `<` not `<=`. Fix:" + +Caveman (ultra): "Auth bug. `<` → `<=`. Fix:" + +## Limits + +* Caveman affects assistant prose only. It does not change generated code, commit messages, or PR descriptions. +* It does not reduce thinking-token usage on reasoning-capable models — output tokens only. + +## Attribution + +Concept based on the [Caveman project](https://github.com/JuliusBrussee/caveman) (MIT license, Copyright (c) 2026 Julius Brussee). This SKILL.md is an original specification authored for hve-core; no upstream files are redistributed. + +> Brought to you by microsoft/hve-core + +🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers. diff --git a/collections/experimental.collection.md b/collections/experimental.collection.md index 046820510..82c20a4f5 100644 --- a/collections/experimental.collection.md +++ b/collections/experimental.collection.md @@ -41,6 +41,7 @@ Experimental and preview artifacts not yet promoted to stable collections. Items | Name | Description | |--------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **caveman** | Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules | | **customer-card-render** | Generate customer-card PowerPoint content YAML from Design Thinking canonical artifacts and build using the shared PowerPoint skill pipeline | | **mural** | Mural workspace, room, mural, and widget workflows via the Mural REST API exposed through a Python CLI. Use when you need to read or write Mural content or automate widget creation. | | **powerpoint** | PowerPoint slide deck generation and management using python-pptx with YAML-driven content and styling | diff --git a/collections/experimental.collection.yml b/collections/experimental.collection.yml index 777e9ef0b..6226a99c0 100644 --- a/collections/experimental.collection.yml +++ b/collections/experimental.collection.yml @@ -14,6 +14,9 @@ items: - path: .github/agents/experimental/subagents/pptx-subagent.agent.md kind: agent # Skills + - path: .github/skills/experimental/caveman + kind: skill + maturity: experimental - path: .github/skills/experimental/customer-card-render kind: skill - path: .github/skills/experimental/powerpoint diff --git a/collections/hve-core-all.collection.yml b/collections/hve-core-all.collection.yml index 4f5cc511d..3cf7306fc 100644 --- a/collections/hve-core-all.collection.yml +++ b/collections/hve-core-all.collection.yml @@ -504,6 +504,9 @@ items: - path: .github/skills/design-thinking/dt-rpi-integration kind: skill maturity: preview +- path: .github/skills/experimental/caveman + kind: skill + maturity: experimental - path: .github/skills/experimental/customer-card-render kind: skill - path: .github/skills/experimental/mural diff --git a/plugins/experimental/README.md b/plugins/experimental/README.md index a0e8a5b07..eb3519383 100644 --- a/plugins/experimental/README.md +++ b/plugins/experimental/README.md @@ -48,6 +48,7 @@ Experimental and preview artifacts not yet promoted to stable collections. Items | Name | Description | |--------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| **caveman** | Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules | | **customer-card-render** | Generate customer-card PowerPoint content YAML from Design Thinking canonical artifacts and build using the shared PowerPoint skill pipeline | | **mural** | Mural workspace, room, mural, and widget workflows via the Mural REST API exposed through a Python CLI. Use when you need to read or write Mural content or automate widget creation. | | **powerpoint** | PowerPoint slide deck generation and management using python-pptx with YAML-driven content and styling | diff --git a/plugins/experimental/skills/experimental/caveman b/plugins/experimental/skills/experimental/caveman new file mode 120000 index 000000000..a93cb3627 --- /dev/null +++ b/plugins/experimental/skills/experimental/caveman @@ -0,0 +1 @@ +../../../../.github/skills/experimental/caveman \ No newline at end of file diff --git a/plugins/hve-core-all/skills/experimental/caveman b/plugins/hve-core-all/skills/experimental/caveman new file mode 120000 index 000000000..a93cb3627 --- /dev/null +++ b/plugins/hve-core-all/skills/experimental/caveman @@ -0,0 +1 @@ +../../../../.github/skills/experimental/caveman \ No newline at end of file From cfb1173de2201ddc746acf4f6d7a4e5c66cc9fef Mon Sep 17 00:00:00 2001 From: Sascha Corti Date: Thu, 18 Jun 2026 08:26:20 +0200 Subject: [PATCH 2/2] feat(caveman): Enhance skill definition with argument hint Adds an `argument-hint` to `SKILL.md` to explicitly list valid usage parameters for the skill. Removes an outdated internal comment and includes minor formatting adjustments in collection and plugin manifests. --- .github/skills/experimental/caveman/SKILL.md | 3 +-- collections/hve-core-all.collection.md | 1 + plugins/hve-core-all/README.md | 1 + 3 files changed, 3 insertions(+), 2 deletions(-) diff --git a/.github/skills/experimental/caveman/SKILL.md b/.github/skills/experimental/caveman/SKILL.md index 80bb2743a..06aefba2f 100644 --- a/.github/skills/experimental/caveman/SKILL.md +++ b/.github/skills/experimental/caveman/SKILL.md @@ -1,6 +1,7 @@ --- name: caveman description: 'Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules - Brought to you by microsoft/hve-core' +argument-hint: "[{lite|full|ultra|wenyan|off}]" license: MIT disable-model-invocation: true metadata: @@ -101,5 +102,3 @@ Caveman (ultra): "Auth bug. `<` → `<=`. Fix:" Concept based on the [Caveman project](https://github.com/JuliusBrussee/caveman) (MIT license, Copyright (c) 2026 Julius Brussee). This SKILL.md is an original specification authored for hve-core; no upstream files are redistributed. > Brought to you by microsoft/hve-core - -🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers. diff --git a/collections/hve-core-all.collection.md b/collections/hve-core-all.collection.md index 53613f55e..b4730314a 100644 --- a/collections/hve-core-all.collection.md +++ b/collections/hve-core-all.collection.md @@ -238,6 +238,7 @@ Use this edition when you want access to everything without choosing a focused c | **adr-author** | Authoring skill for Architecture Decision Records (ADRs) supporting capture, from-planner-handoff, and adopt-template entry modes with selectable Y-Statement or MADR v4.0.0 output templates, supersession lineage, and ASR trigger evaluation - Brought to you by microsoft/hve-core. | | **architecture-diagrams** | Architecture diagram authoring for cloud infrastructure: parse Azure IaC, map relationships, and render either ASCII block diagrams or Mermaid flowcharts based on the caller's chosen output format | | **backlog-templates** | Shared work-item templates and conventions for ADO and GitHub backlog handoff across the RAI, Security, SSSC, and Accessibility planners | +| **caveman** | Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules | | **customer-card-render** | Generate customer-card PowerPoint content YAML from Design Thinking canonical artifacts and build using the shared PowerPoint skill pipeline | | **documentation** | Canonical documentation capability for audit, drift, validate, and author modes in hve-core. | | **dt-coaching-foundation** | Design Thinking coaching foundation knowledge: coach identity and philosophy, quality and fidelity constraints, method sequencing, coaching state schema, and the canonical deck workflow | diff --git a/plugins/hve-core-all/README.md b/plugins/hve-core-all/README.md index 18a1943ec..9dd6a8d95 100644 --- a/plugins/hve-core-all/README.md +++ b/plugins/hve-core-all/README.md @@ -243,6 +243,7 @@ Use this edition when you want access to everything without choosing a focused c | **adr-author** | Authoring skill for Architecture Decision Records (ADRs) supporting capture, from-planner-handoff, and adopt-template entry modes with selectable Y-Statement or MADR v4.0.0 output templates, supersession lineage, and ASR trigger evaluation - Brought to you by microsoft/hve-core. | | **architecture-diagrams** | Architecture diagram authoring for cloud infrastructure: parse Azure IaC, map relationships, and render either ASCII block diagrams or Mermaid flowcharts based on the caller's chosen output format | | **backlog-templates** | Shared work-item templates and conventions for ADO and GitHub backlog handoff across the RAI, Security, SSSC, and Accessibility planners | +| **caveman** | Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules | | **customer-card-render** | Generate customer-card PowerPoint content YAML from Design Thinking canonical artifacts and build using the shared PowerPoint skill pipeline | | **documentation** | Canonical documentation capability for audit, drift, validate, and author modes in hve-core. | | **dt-coaching-foundation** | Design Thinking coaching foundation knowledge: coach identity and philosophy, quality and fidelity constraints, method sequencing, coaching state schema, and the canonical deck workflow |