# caveman: what changed from 1 to 2

- Product: caveman (https://whatsnew.fyi/product/caveman)
- Vendor: Julius Brussee
- Range: changelog entries numbered after v1.10.0 up to and including v2.6.0, stable releases only
- Entries below: 8 releases (newest first)
- Resolved: 1 is v1.10.0 and 2 is v2.6.0, the newest stable release of each major we track
- Not shown: 8 releases whose versions could not be placed against this range
- Carrying security changes: 2 · CVEs mentioned: 0 · Mentioning breaking changes: 0 · Removing or deprecating something: 3
- Page: https://whatsnew.fyi/product/caveman/compare/1...2

What's New is an index, not a publisher: every entry below links to the vendor's own release notes, which are the authoritative source. Entries are labelled where they are hand-curated sample data, pre-releases, or drawn from a secondary source such as a developer blog.

Reuse: the summaries, labels and curation here are © What's New. Quote freely with attribution and a link back; wholesale republication of the corpus is not permitted — terms: https://whatsnew.fyi/terms. The vendors' own release notes remain their publishers'.

## What changed (85 changes, grouped by kind)

### Added

#### v2.6.0 (2026-09-04)

- Firefox build of the Caveman Mode extension with AMO-clean manifest and addons-linter pass
- ASD-STE100 clarity register folded into caveman skill to keep compressed answers unambiguous while maintaining caveman voice
- Per-model originals and compat-mount routing for custom providers
- Real-Firefox injection guard behind npm run test:firefox

#### v2.5.0 (2026-09-02)

- Qwen Code and Kilo Code wraps refuse extension, ACP, session-restore, control-command, daemon, and console surfaces that could carry model traffic outside the locked route
- Standalone proxies prove MCP recovery from the request's own tool list, so subscription sessions compress without a wrap
- Claude Code Remote Control is documented as incompatible with a proxied base URL and launches direct under wrap

#### v2.3.1 (2026-08-23)

- caveman learn now prices the scanned window from provider-counted usage read out of agent transcripts
- Confirmed savings are attributed to a named method: deterministic_remeasure, controlled_holdout, counterfactual_replay, or interrupted_time_series
- Controlled-holdout harness runs a fix on and off on the same machine and returns insufficient_data under 5 sessions per arm
- caveman learn reconcile checks measured spend against a provider usage export with a privacy-safe export digest
- Test asserts that bootstrap pins, installer package version, and documented one-liners all name the same release

#### v2.3.0 (2026-08-23)

- Confirmed savings are attributed to a named method — deterministic_remeasure, controlled_holdout, counterfactual_replay, or interrupted_time_series
- A controlled-holdout harness runs a fix on and off on the same machine to judge changes whose payback lands outside the edited file
- caveman learn reconcile checks measured spend against a provider usage export and generates a privacy-safe export digest

#### v2.2.0 (2026-08-20)

- benchmarks/run.py gains a terse control arm to separate skill improvements from baseline instruction effects

#### v2.1.0 (2026-08-16)

- Context-depth report showing each session's peak context-window share, bucketed, with over-30% and over-50% counts
- Report UI with TLDR summary, two-card savings display showing proxy-measured saved-so-far and replay could-have-saved, and 30-day list-price cost illustration
- Persisted shortcut and token telemetry in CLI
- New twenty-chapter technical manual under docs/technical/

#### v2.0.0 (2026-08-11)

- Add a compression engine that compresses tool outputs, logs, code, and JSON before sending to the model
- Add a proxy component to integrate compression into LLM applications
- Add CLI tool for compression functionality
- Add SDKs for multiple programming languages
- Add compression evaluation tools and contracts for measuring compression savings

### Changed

#### v2.6.0 (2026-09-04)

- Array-shaped tool output is now compressed
- Route eligibility is now recorded
- Stronger per-turn caveman reinforcement so competing mid-conversation instructions do not drift the style
- Native-hook bridge returns on the first complete JSON payload instead of waiting for EOF
- CLI's host hook callbacks apply same first-JSON rule
- Number formatting pinned to en-US
- Only the default provider's endpoints are routed for custom providers

#### v2.5.0 (2026-09-02)

- Provider-native cache optimizers and the breakpoint planner default on in optimization modes
- README rewritten around measured savings with static charts rendered from the committed benchmark table

#### v2.3.1 (2026-08-23)

- Spend is reported as the actual window cost, never projected into a month or called verified
- Proxy gateway buffers non-streaming upstream bodies before replying and serve-mode warnings land in ~/.caveman/proxy.log
- Pinned bin-v1.1.3 companions and CLI 1.2.4

#### v2.3.0 (2026-08-23)

- caveman learn now reports the actual cost of the window from provider-counted usage instead of just token movements
- The gateway buffers non-streaming upstream bodies before replying
- The caveman ruleset's framing clauses were rewritten and the bundled agent skill registry was regenerated

#### v2.2.0 (2026-08-20)

- cavecrew model overrides no longer rewrite tracked agents/*.md when the plugin root is a source checkout

#### v2.1.0 (2026-08-16)

- README rewritten as a landing page with proxy-first install, caveman learn showcase with real report screenshot, and skills-as-images
- Skill forbids caveman phrasing that adds tokens and names defect/ticket text
- Hooks resolve punctuated /caveman levels and stop prose firing mode triggers
- Installer installs Cursor skills globally
- Opencode agents drop provider-less model values

#### v2.0.0 (2026-08-11)

- Change licensing model with MIT for skill, CLI, SDKs, kit, evals, contracts, catalog, and extension shell; BSL-1.1 for engine, proxy, MCP, shrink, cavemem-core, and platform with sunset to Apache-2.0 in approximately 4 years

### Fixed

#### v2.6.0 (2026-09-04)

- Routing bug
- Hook bug
- Compression bug
- Config self-corruption in wrap
- Duplicate hooks stacking
- x-api-key is sent on compat /v1/messages
- Dropped indicator font-face that 404'd on every chat page
- Cavecrew subagents dropped the Claude-only tools frontmatter so Gemini CLI loads them
- Freshly started proxy's run state is awaited before judging its owner

#### v2.5.0 (2026-09-02)

- caveman-compress validates each candidate in a staging file before touching the live file and tolerates a leading tool_use block
- caveman-stats counts each multi-block response once

#### v2.4.0 (2026-08-29)

- Scope repo intelligence to project source and require direct evidence in proxy
- Handle stdout/stderr 'error' in UserPromptSubmit hook
- Honor HTTP_PROXY/HTTPS_PROXY for outbound requests in CLI
- Block sensitive directory names in compression
- Add Claude 5 model output pricing to stats
- Scope caveman mode to the session
- Avoid CLI bin collision in installer
- Serialize concurrent compress runs with a cross-session lock
- Support documented version flag in CLI

#### v2.3.1 (2026-08-23)

- Installer pins in install.sh and install.ps1 were out of sync with bin/install.js and the documented one-liners, causing the v2.3.0 one-liner to bootstrap the previous release

#### v2.3.0 (2026-08-23)

- Serve-mode warnings now land in ~/.caveman/proxy.log instead of vanishing

#### v2.2.0 (2026-08-20)

- Uninstall no longer deletes hook files while settings.json still points at them
- Settings file that cannot be parsed or rewritten now causes a hard stop
- POSIX and PowerShell uninstallers exit non-zero on missing node
- Install leaves another plugin's hooks/package.json alone and removes only the manifest caveman ships
- Binary installers give each download its own .part file to prevent concurrent installs from publishing each other's half-written files
- Compress skill outer-fence strip no longer merges a document's first and last code blocks into prose
- Compress skill validator now sees CommonMark indented code blocks
- Compress skill treats lost file paths and renamed headings as errors instead of warnings

#### v2.1.0 (2026-08-16)

- Hooks degrade instead of crashing when a sibling hook file is missing
- Compress: indented fence markers no longer break inline-code validation
- SSRF block errors now name the allowlist escape hatch

### Removed

#### v2.5.0 (2026-09-02)

- cacheengine/ removed from this repository

#### v2.4.0 (2026-08-29)

- Remove deprecated hooks alias from codex

#### v2.3.0 (2026-08-23)

- The GitHub Pages landing page is gone

### Security

#### v2.5.0 (2026-09-02)

- Wrap that finds a listener on the proxy port it did not start and cannot verify now launches the agent directly, so provider keys never go to a foreign process

#### v2.2.0 (2026-08-20)

- Installer temp downloads move to a 0700 mkdtemp to prevent a pre-planted symlink from capturing the init script

## Release notes

### v2.6.0

- Date: 2026-09-04
- Version: v2.6.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.6.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.6.0

- **added** — Firefox build of the Caveman Mode extension with AMO-clean manifest and addons-linter pass
- **added** — ASD-STE100 clarity register folded into caveman skill to keep compressed answers unambiguous while maintaining caveman voice
- **added** — Per-model originals and compat-mount routing for custom providers
- **added** — Real-Firefox injection guard behind npm run test:firefox
- **changed** — Array-shaped tool output is now compressed
- **changed** — Route eligibility is now recorded
- **changed** — Stronger per-turn caveman reinforcement so competing mid-conversation instructions do not drift the style
- **changed** — Native-hook bridge returns on the first complete JSON payload instead of waiting for EOF
- **changed** — CLI's host hook callbacks apply same first-JSON rule
- **changed** — Number formatting pinned to en-US
- **changed** — Only the default provider's endpoints are routed for custom providers
- **fixed** — Routing bug
- **fixed** — Hook bug
- **fixed** — Compression bug
- **fixed** — Config self-corruption in wrap
- **fixed** — Duplicate hooks stacking
- **fixed** — x-api-key is sent on compat /v1/messages
- **fixed** — Dropped indicator font-face that 404'd on every chat page
- **fixed** — Cavecrew subagents dropped the Claude-only tools frontmatter so Gemini CLI loads them
- **fixed** — Freshly started proxy's run state is awaited before judging its owner

Caveman 2.6.0 ships a Firefox build of the Caveman Mode extension, folds an ASD-STE100 clarity register into the caveman skill, and fixes the routing, hook, and compression bugs found since 2.5.0.

Highlights:
- Skill: caveman prose now always mixes in an ASD-STE100 clarity register, so compressed answers stay unambiguous without dropping out of caveman voice.
- Extension: Firefox port of the Caveman Mode extension (#810), AMO-clean manifest and addons-linter pass, a Firefox pack target in the zip builder and the release gate, and a real-Firefox injection guard behind `npm run test:firefox`. Dropped an indicator font-face that 404'd on every chat page.
- Codex: array-shaped tool output is compressed, route eligibility is recorded, and the wrap no longer self-corrupts its config or stacks duplicate hooks.
- Pi: per-model originals and compat-mount routing for custom providers; only the default provider's endpoints are routed (#972).
- Proxy: `x-api-key` is sent on compat `/v1/messages` (#969), and the native-hook bridge returns on the first complete JSON payload instead of waiting for EOF (#965). Same first-JSON rule applied to the CLI's host hook callbacks.
- Hooks: stronger per-turn caveman reinforcement so competing mid-conversation instructions do not drift the style (#951).
- Agents: cavecrew subagents dropped the Claude-only `tools` frontmatter so Gemini CLI loads them (#966).
- Stats: number formatting pinned to en-US (#954).
- CLI: a freshly started proxy's run state is awaited before judging its owner.

Artifacts:
- Signed binaries: bin-v1.1.6 (built by CI from that tag)
- CLI: `npm install -g @caveman-ai/cli@1.3.3` (npm publish pending)

Validation:
- `tests/verify_repo.py`: all local verification checks passed.
- Go proxy suite (`go test ./...` in `proxy/`): green.
- Installer bootstrap pin lockstep test: green.

### v2.5.0

- Date: 2026-09-02
- Version: v2.5.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.5.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.5.0

- **security** — Wrap that finds a listener on the proxy port it did not start and cannot verify now launches the agent directly, so provider keys never go to a foreign process
- **added** — Qwen Code and Kilo Code wraps refuse extension, ACP, session-restore, control-command, daemon, and console surfaces that could carry model traffic outside the locked route
- **added** — Standalone proxies prove MCP recovery from the request's own tool list, so subscription sessions compress without a wrap
- **changed** — Provider-native cache optimizers and the breakpoint planner default on in optimization modes
- **fixed** — caveman-compress validates each candidate in a staging file before touching the live file and tolerates a leading tool_use block
- **fixed** — caveman-stats counts each multi-block response once
- **added** — Claude Code Remote Control is documented as incompatible with a proxied base URL and launches direct under wrap
- **changed** — README rewritten around measured savings with static charts rendered from the committed benchmark table
- **removed** — cacheengine/ removed from this repository

Caveman 2.5.0 closes this week's issue and PR queue, drops the standalone cache engine from the repository, and ships a fail-closed fix for proxy-port takeover.

Highlights:
- Security: a wrap that finds a listener on the proxy port it did not start and cannot verify now launches the agent directly, so provider keys never go to a foreign process (#945).
- Qwen Code and Kilo Code wraps refuse extension, ACP, session-restore, control-command, daemon, and console surfaces that could carry model traffic outside the locked route.
- Standalone proxies prove MCP recovery from the request's own tool list, so subscription sessions compress without a wrap (#908).
- Provider-native cache optimizers and the breakpoint planner default on in optimization modes; explicit off still wins.
- caveman-compress validates each candidate in a staging file before touching the live file (#544, #938) and tolerates a leading tool_use block.
- caveman-stats counts each multi-block response once (#793, #794).
- Claude Code Remote Control is documented as incompatible with a proxied base URL and launches direct under wrap (#947).
- README rewritten around measured savings with static charts rendered from the committed benchmark table.
- cacheengine/ removed from this repository.

Artifacts:
- CLI: `npm install -g @caveman-ai/cli@1.3.2` (npm publish pending)
- Signed binaries: bin-v1.1.5 (built by CI from this tag)

Validation:
- Full CLI suite locally: 943 tests, 942 passed, 18 skipped; the one failure is the raw-wrap base-URL test on a machine whose own caveman proxy already holds port 8787, which the new fail-closed path treats as unverifiable.
- Go proxy suite green except the pre-existing learn_retro store failures, unchanged from v2.4.0.
- verify_repo, Python, hook, and installer pin tests green.

### v2.4.0

- Date: 2026-08-29
- Version: v2.4.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.4.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.4.0

- **fixed** — Scope repo intelligence to project source and require direct evidence in proxy
- **fixed** — Handle stdout/stderr 'error' in UserPromptSubmit hook
- **fixed** — Honor HTTP_PROXY/HTTPS_PROXY for outbound requests in CLI
- **fixed** — Block sensitive directory names in compression
- **fixed** — Add Claude 5 model output pricing to stats
- **fixed** — Scope caveman mode to the session
- **removed** — Remove deprecated hooks alias from codex
- **fixed** — Avoid CLI bin collision in installer
- **fixed** — Serialize concurrent compress runs with a cross-session lock
- **fixed** — Support documented version flag in CLI

Caveman 2.4.0 closes the full weekly issue and PR queue with adversarial review, targeted fixes, and release-grade validation.

Highlights:
- Hardened proxy redirect, authentication, cancellation, and TLS behavior.
- Request-scoped MCP recovery and per-session native-runtime locking.
- Safer Windows hooks and lifecycle handling.
- Stronger compression masking, isolation, accounting, and semantic doctor checks.
- Integrated accepted contributor work with explicit attribution and closure messages.
- Restored documented `caveman --version` behavior.

Artifacts:
- CLI: `npm install -g @caveman-ai/cli@1.3.1`
- Signed binaries: [bin-v1.1.4](https://github.com/JuliusBrussee/caveman/releases/tag/bin-v1.1.4)

Validation:
- Full CLI suite: 665 tests, 647 passed, 18 skipped, 0 failed.
- Required Linux, macOS, Windows, Go, TypeScript, Python, extension, and root CI passed.
- Clean npm install, signed-binary checksum verification, direct wrap, compression, and byte-exact retrieval smoke passed.
##### What's Changed
* fix(proxy): scope repo intelligence to project source and require direct evidence by @JuliusBrussee in https://github.com/JuliusBrussee/caveman/pull/901
* fix: harden weekly integrations and prepare v2.4.0 by @JuliusBrussee in https://github.com/JuliusBrussee/caveman/pull/929
* fix(hooks): handle stdout/stderr 'error' in UserPromptSubmit hook by @AmirF194 in https://github.com/JuliusBrussee/caveman/pull/795
* fix(cli): honour HTTP_PROXY/HTTPS_PROXY for outbound requests by @dylanneve1 in https://github.com/JuliusBrussee/caveman/pull/889
* fix(compress): block sensitive directory names by @CooperSheroy in https://github.com/JuliusBrussee/caveman/pull/891
* chore(skills): shorten skill descriptions to fit the listing budget by @smoochy in https://github.com/JuliusBrussee/caveman/pull/898
* fix(stats): add Claude 5 model output pricing by @SomSamantray in https://github.com/JuliusBrussee/caveman/pull/900
* fix: scope caveman mode to the session by @pavel-collab in https://github.com/JuliusBrussee/caveman/pull/904
* fix(codex): remove deprecated hooks alias by @cuddy-five in https://github.com/JuliusBrussee/caveman/pull/927
* fix(installer): avoid CLI bin collision by @pttydou in https://github.com/JuliusBrussee/caveman/pull/928
* fix(compress): serialize concurrent compress runs with a cross-session lock by @hunleyd in https://github.com/JuliusBrussee/caveman/pull/798
* fix(cli): support documented version flag by @JuliusBrussee in https://github.com/JuliusBrussee/caveman/pull/930

##### New Contributors
* @dylanneve1 made their first contribution in https://github.com/JuliusBrussee/caveman/pull/889
* @CooperSheroy made their first contribution in https://github.com/JuliusBrussee/caveman/pull/891
* @smoochy made their first contribution in https://github.com/JuliusBrussee/caveman/pull/898
* @SomSamantray made their first contribution in https://github.com/JuliusBrussee/caveman/pull/900
* @pavel-collab made their first contribution in https://github.com/JuliusBrussee/caveman/pull/904
* @cuddy-five made their first contribution in https://github.com/JuliusBrussee/caveman/pull/927
* @pttydou made their first contribution in https://github.com/JuliusBrussee/caveman/pull/928
* @hunleyd made their first contribution in https://github.com/JuliusBrussee/caveman/pull/798

**Full Changelog**: https://github.com/JuliusBrussee/caveman/compare/v2.3.1...v2.4.0

### v2.3.1

- Date: 2026-08-23
- Version: v2.3.1
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.3.1
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.3.1

- **fixed** — Installer pins in install.sh and install.ps1 were out of sync with bin/install.js and the documented one-liners, causing the v2.3.0 one-liner to bootstrap the previous release
- **added** — caveman learn now prices the scanned window from provider-counted usage read out of agent transcripts
- **added** — Confirmed savings are attributed to a named method: deterministic_remeasure, controlled_holdout, counterfactual_replay, or interrupted_time_series
- **added** — Controlled-holdout harness runs a fix on and off on the same machine and returns insufficient_data under 5 sessions per arm
- **added** — caveman learn reconcile checks measured spend against a provider usage export with a privacy-safe export digest
- **changed** — Spend is reported as the actual window cost, never projected into a month or called verified
- **changed** — Proxy gateway buffers non-streaming upstream bodies before replying and serve-mode warnings land in ~/.caveman/proxy.log
- **added** — Test asserts that bootstrap pins, installer package version, and documented one-liners all name the same release
- **changed** — Pinned bin-v1.1.3 companions and CLI 1.2.4

Everything in [v2.3.0](https://github.com/JuliusBrussee/caveman/releases/tag/v2.3.0) — learn v3 spend and attributed savings — plus the installer-pin fix below. **Install v2.3.1, not v2.3.0.**

##### Installer pin fix

v2.3.0 bumped `bin/install.js` and the documented one-liners but left `install.sh` and `install.ps1` pinned at `v2.2.0`. The curl-pipe path reads the shim's own default ref, so the v2.3.0 one-liner bootstrapped the previous release. The root installer package version had drifted the same way.

A test now asserts that the three bootstrap pins, the installer package version, and every documented one-liner name the same release, so the next bump cannot miss one silently.

##### Learn v3 (from v2.3.0)

- `caveman learn` prices the scanned window from provider-counted usage read out of agent transcripts, so the cost figure is the provider's arithmetic rather than an estimate.
- Confirmed savings are attributed to a **named method** — `deterministic_remeasure`, `controlled_holdout`, `counterfactual_replay`, or `interrupted_time_series` — and rungs are never summed into one blended headline.
- A **controlled-holdout harness** runs a fix on and off on the same machine; under 5 sessions per arm it returns `insufficient_data` rather than a small win.
- `caveman learn reconcile` checks measured spend against a provider usage export, and a privacy-safe export digest makes that export shareable without shipping prompt text.
- Spend is what the window cost — never what a fix would return, never projected into a month, never called verified.
- Proxy: the gateway buffers non-streaming upstream bodies before replying, and serve-mode warnings land in `~/.caveman/proxy.log`.

##### Runtime

Pins the **bin-v1.1.3** companions and **CLI 1.2.4**. All 36 binaries and the signed manifest are at [bin-v1.1.3](https://github.com/JuliusBrussee/caveman/releases/tag/bin-v1.1.3).

##### Install

```bash
curl -fsSL https://raw.githubusercontent.com/JuliusBrussee/caveman/v2.3.1/install.sh | bash
```

```powershell
irm https://raw.githubusercontent.com/JuliusBrussee/caveman/v2.3.1/install.ps1 | iex
```

### v2.3.0

- Date: 2026-08-23
- Version: v2.3.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.3.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.3.0

- **changed** — caveman learn now reports the actual cost of the window from provider-counted usage instead of just token movements
- **added** — Confirmed savings are attributed to a named method — deterministic_remeasure, controlled_holdout, counterfactual_replay, or interrupted_time_series
- **added** — A controlled-holdout harness runs a fix on and off on the same machine to judge changes whose payback lands outside the edited file
- **added** — caveman learn reconcile checks measured spend against a provider usage export and generates a privacy-safe export digest
- **changed** — The gateway buffers non-streaming upstream bodies before replying
- **fixed** — Serve-mode warnings now land in ~/.caveman/proxy.log instead of vanishing
- **changed** — The caveman ruleset's framing clauses were rewritten and the bundled agent skill registry was regenerated
- **removed** — The GitHub Pages landing page is gone

> **Superseded by [v2.3.1](https://github.com/JuliusBrussee/caveman/releases/tag/v2.3.1).** This tag's `install.sh` / `install.ps1` are still pinned at v2.2.0, so the v2.3.0 one-liner bootstraps the previous release. Use v2.3.1.

Learn v3: `caveman learn` now says what the window **cost**, not just how many tokens moved — and every savings number it reports carries the method that measured it.

##### Learn v3 — spend and attributed savings

- The scan reads billing buckets and outcome signals straight from agent transcripts (Claude, Codex, Gemini, opencode) and prices the window from provider-counted usage, so the cost figure is the provider's arithmetic rather than an estimate.
- Confirmed savings are attributed to a **named method** — `deterministic_remeasure`, `controlled_holdout`, `counterfactual_replay`, or `interrupted_time_series`. Rungs are never summed into one blended headline, because a re-counted file and a before/after median are not the same kind of evidence.
- A **controlled-holdout harness** runs a fix on and off on the same machine, so a change whose payback lands somewhere other than the file it edited (a distilled skill, say) can still be judged. Under 5 sessions per arm it returns `insufficient_data` instead of a small win.
- `caveman learn reconcile` checks measured spend against a provider usage export, and a privacy-safe **export digest** makes that export shareable without shipping prompt text.
- Spend, the savings ledger, and holdout reports surface in the CLI and in the `caveman-learn` skill. Spend is what the window cost — never what a fix would return, never projected into a month, and never called verified. On a Max/Plus plan it is the API-equivalent value of the tokens, not money spent.

##### Proxy

- The gateway buffers non-streaming upstream bodies before replying.
- Serve-mode warnings land in `~/.caveman/proxy.log` instead of vanishing.

##### Elsewhere

- The caveman ruleset's framing clauses were rewritten, and the bundled agent skill registry regenerated.
- The GitHub Pages landing page is gone.

##### Runtime

Installs pin the **bin-v1.1.3** companions and **CLI 1.2.4**. All 36 binaries and the signed manifest are published at [bin-v1.1.3](https://github.com/JuliusBrussee/caveman/releases/tag/bin-v1.1.3).

##### Install

```bash
curl -fsSL https://raw.githubusercontent.com/JuliusBrussee/caveman/v2.3.0/install.sh | bash
```

```powershell
irm https://raw.githubusercontent.com/JuliusBrussee/caveman/v2.3.0/install.ps1 | iex
```

### v2.2.0

- Date: 2026-08-20
- Version: v2.2.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.2.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.2.0

- **fixed** — Uninstall no longer deletes hook files while settings.json still points at them
- **fixed** — Settings file that cannot be parsed or rewritten now causes a hard stop
- **fixed** — POSIX and PowerShell uninstallers exit non-zero on missing node
- **fixed** — Install leaves another plugin's hooks/package.json alone and removes only the manifest caveman ships
- **security** — Installer temp downloads move to a 0700 mkdtemp to prevent a pre-planted symlink from capturing the init script
- **fixed** — Binary installers give each download its own .part file to prevent concurrent installs from publishing each other's half-written files
- **fixed** — Compress skill outer-fence strip no longer merges a document's first and last code blocks into prose
- **fixed** — Compress skill validator now sees CommonMark indented code blocks
- **fixed** — Compress skill treats lost file paths and renamed headings as errors instead of warnings
- **changed** — cavecrew model overrides no longer rewrite tracked agents/*.md when the plugin root is a source checkout
- **added** — benchmarks/run.py gains a terse control arm to separate skill improvements from baseline instruction effects

Safety fixes across the installer, the uninstaller, and the compress skill.

##### Install / uninstall

- Uninstall no longer deletes hook files while `settings.json` still points at them — that is #471 on every session start. A settings file it cannot parse or rewrite is now a hard stop, and the POSIX and PowerShell uninstallers exit non-zero on a missing `node` for the same reason.
- Install leaves another plugin's `hooks/package.json` alone; uninstall removes only the manifest caveman ships.
- Installer temp downloads move to a `0700` mkdtemp, so a pre-planted symlink can no longer capture the init script the installer then executes.
- The binary installers give each download its own `.part`. The fixed name made the `O_EXCL` guard unreachable, letting concurrent installs publish each other's half-written files as "checksum verified".

##### Compress skill

- The outer-fence strip no longer merges a document's first and last code blocks into prose. It matched any file that merely started and ended with a fence line.
- The validator now sees CommonMark indented code blocks — a 4-space-indented `kubectl delete` was prose to it, so a rewritten destructive command passed clean and overwrote the file.
- Lost file paths and renamed headings are errors, not warnings that let the in-place overwrite stand.

##### Hooks

- cavecrew model overrides no longer rewrite tracked `agents/*.md` when the plugin root is a source checkout.

##### Benchmarks

- `benchmarks/run.py` gains a terse control arm, so the next regenerated table separates what the skill adds from what any "answer concisely" instruction already buys. The published 65% is against an unprompted baseline and now says so.

##### Install

```bash
curl -fsSL https://raw.githubusercontent.com/JuliusBrussee/caveman/v2.2.0/install.sh | bash
```

```powershell
irm https://raw.githubusercontent.com/JuliusBrussee/caveman/v2.2.0/install.ps1 | iex
```

### v2.1.0

- Date: 2026-08-16
- Version: v2.1.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.1.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.1.0

- **added** — Context-depth report showing each session's peak context-window share, bucketed, with over-30% and over-50% counts
- **added** — Report UI with TLDR summary, two-card savings display showing proxy-measured saved-so-far and replay could-have-saved, and 30-day list-price cost illustration
- **added** — Persisted shortcut and token telemetry in CLI
- **added** — New twenty-chapter technical manual under docs/technical/
- **changed** — README rewritten as a landing page with proxy-first install, caveman learn showcase with real report screenshot, and skills-as-images
- **changed** — Skill forbids caveman phrasing that adds tokens and names defect/ticket text
- **changed** — Hooks resolve punctuated /caveman levels and stop prose firing mode triggers
- **fixed** — Hooks degrade instead of crashing when a sibling hook file is missing
- **fixed** — Compress: indented fence markers no longer break inline-code validation
- **changed** — Installer installs Cursor skills globally
- **fixed** — SSRF block errors now name the allowlist escape hatch
- **changed** — Opencode agents drop provider-less model values

##### Caveman 2.1.0

###### caveman learn
- Context-depth report: each session's peak context-window share, bucketed, with over-30%/over-50% counts.
- Report UI: TLDR summary, two-card savings (proxy-measured saved-so-far + replay could-have-saved), 30-day list-price cost illustration.

###### CLI
- Persisted shortcut and token telemetry (feat/cli).

###### Skill and hooks
- Skill forbids caveman phrasing that adds tokens; names defect/ticket text.
- Hooks resolve punctuated /caveman levels and stop prose firing mode triggers; degrade instead of crashing when a sibling hook file is missing.
- Compress: indented fence markers no longer break inline-code validation.
- Installer installs Cursor skills globally; SSRF block errors name the allowlist escape hatch; opencode agents drop provider-less model values.

###### Docs
- New twenty-chapter technical manual under docs/technical/.
- README rewritten as a landing page: proxy-first install, caveman learn showcase with a real report screenshot, skills-as-images.

All local numbers remain `inferred`; benchmark numbers remain `benchmark_counterfactual`. Installer pins bumped to v2.1.0.

### v2.0.0 — Caveman 2.0.0 — the engine ships

- Date: 2026-08-11
- Version: v2.0.0
- Original notes: https://github.com/JuliusBrussee/caveman/releases/tag/v2.0.0
- Permalink: https://whatsnew.fyi/product/caveman/releases/v2.0.0

- **added** — Add a compression engine that compresses tool outputs, logs, code, and JSON before sending to the model
- **added** — Add a proxy component to integrate compression into LLM applications
- **added** — Add CLI tool for compression functionality
- **added** — Add SDKs for multiple programming languages
- **changed** — Change licensing model with MIT for skill, CLI, SDKs, kit, evals, contracts, catalog, and extension shell; BSL-1.1 for engine, proxy, MCP, shrink, cavemem-core, and platform with sunset to Apache-2.0 in approximately 4 years
- **added** — Add compression evaluation tools and contracts for measuring compression savings

**Caveman 2 — the engine ships.** v1 shrank the model's mouth (output). v2 adds the ears: a real compression engine that squeezes what the model *reads* — tool outputs, logs, code, JSON — before it ever hits the model. One line in front of any LLM app. Free, local, BYOK.

**The skill is untouched and still MIT.** Nothing about your `/caveman` install changed.

Real run: a 40-record JSON went `16,098 → 1,091 tokens` (93% smaller), byte-safe, fully recoverable via CCR. Savings are **inferred** — measured on your machine, not projected, not "verified."

**Licensing:** MIT for the skill, CLI, SDKs, kit, evals, contracts, catalog, and extension shell. BSL-1.1 for the engine/proxy/mcp/shrink/cavemem-core/platform — self-host your own traffic (incl. production) free; only reselling Caveman-as-a-service needs a commercial license. Sunsets to Apache-2.0 in ~4 years.

**Ships now:** engine, proxy, CLI, SDKs (as the packages resolve). **Coming soon:** browser extension. Install commands and pip/npm/docker badges go live when the packages publish.

Want to help? New content-type compressors are the highest-leverage contribution — check `good first issue`.