Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/LucasDuys/forgenpx agentmods add skills/lucasduys/forge/caveman-internalWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/lucasduys/forge/caveman-internal)<a href="https://agentmods.dev/skills/lucasduys/forge/caveman-internal"><img src="https://agentmods.dev/badge/skills/lucasduys/forge/caveman-internal/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/lucasduys/forge/caveman-internal"><img src="https://agentmods.dev/badge/skills/lucasduys/forge/caveman-internal.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00036 | $0.01817 |
| Opus 5.5 | $0.00014 | $0.00727 |
| Sonnet 5.5 | $0.00007 | $0.00363 |
| Haiku 4.5 | $0.00004 | $0.00182 |
Grade A, and why
caveman-internal scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Caveman Internal Skill
Compress prose in Forge's own agent-to-agent artifacts so the autonomous loop spends fewer tokens on coordination and more on actual implementation. This skill is consumed internally by Forge agents (executor, planner, reviewer, verifier) when writing handoff notes, artifact summaries, and review reports. It is NOT a user-facing command and must never be invoked as /caveman.
Scope
In scope (compress these):
.forge/artifacts/{task-id}.jsonartifact descriptions and key_decisions.forge/context-bundles/handoff notes between tasks- Review summaries written by the reviewer agent
- Internal status reports passed between agents
- Token-ledger annotations and state.md "What's Done" entries
Out of scope (ALWAYS use normal verbose mode regardless of intensity):
- All code output (source files, code blocks, diffs, patches)
- Commit messages and PR descriptions
- User-facing spec files in
.forge/specs/ - Security warnings of any kind
- Irreversible action confirmations (deletes, force-pushes, schema drops)
- Error messages humans need to act on
- Anything the user will read directly
Intensity Modes
Three levels. Higher intensity removes more, but technical terms (function names, file paths, identifiers, version numbers, error codes) stay verbatim at every level.
lite
Strip filler words and pleasantries. Keep articles, conjunctions, and full grammar. This is the safest mode and is the default for standard depth tasks.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2mo ago First seen · 161 lines · 36 tokens per session scan A 1f2dc8420230
caveman-internal is a skill published in the GitHub repository LucasDuys/forge (58 stars, last pushed 3mo ago), licensed MIT. It adds 36 tokens to every session and 1,817 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
learn
Analyze conversation and update CLAUDE.md, rule files, and/or any skills, commands, or agents that were used — based on user feedback vs agent behavior. Use this skill at the end of any conversation where corrections, misunderstandings, or new patterns emerged, especially when a skill, command, or agent underperformed…
ts-debug
TypeScript/Node debugging expert. Use when the user needs to debug, profile, or trace TypeScript or Node.js code — e.g. "how do I debug this", "find the memory leak", "why is this slow", "add a breakpoint", "profile this function", "why won't the process exit".
plugin-browser
Browse and discover Claude Code skills, agents, commands, and plugins from community and official repositories. Use when the user wants to find, explore, or learn about existing skills, agents, commands, or plugins — e.g., "what plugins are available?", "is there a skill for code review?", "show me community agents"…
py-debug
Python debugging expert. Use when the user needs to debug, profile, or trace Python code — e.g. "how do I debug this", "find the memory leak", "why is this slow", "add a breakpoint", "profile this function".
agent-harness
Browse and explore agent harness frameworks and Claude Code resource collections. Use when the user wants to learn about, explore, or use agent orchestration frameworks, multi-agent architectures, or Claude Code resource catalogs — e.g., "how does Archon work?", "show me everything-claude-code resources", "find an…
prompt-engineering
Expert prompt engineering guidance with proven techniques and anti-patterns. Invoke this skill whenever writing, editing, reviewing, or refining any prompt that will be interpreted by an AI model — system prompts, agent instructions, tool-use prompts, command templates, multi-step task prompts, or agentic workflows.…