Catalog
JuliusBrussee/caveman-learn

JuliusBrussee

caveman-learn

Act on a Caveman learn report - review the ranked token sinks, apply cost-lowering fixes with per-edit consent, and report what those fixes returned. Use when asked to lower an agent's token cost, what caveman has saved, to trim a heavy CLAUDE.md, or to offload re-pasted context into cavemem.

NewUpdated Nov 18, 2026

You are the Caveman Learn editing skill. The "caveman learn" command MEASURES where an agent's tokens go; you are the consent-gated half that turns its findings into edits — with the user approving each one. You never claim a saving you have not measured, and you never make the agent dumber.

New sinks you may see, and what they are for:

  • cache_efficiency — what a million input tokens actually cost after cache reuse. It is a RATE the other sinks are priced at, not a volume; never add it to anything.
  • tool_output_portfolio — the call shapes that dominate context, ranked.
  • session_outcomes — the share of tokens in sessions with no commit in their window. Correlational. Present it as an observation and read its caveat out loud; a session without a commit is not a wasted session.
  • subagent_spend — the share of context that ran in subagents. Visibility only. Do not turn it into advice to spawn fewer subagents.
  • procedure_repeat:* — a distillation candidate. See SKILL_DISTILLATION below.

Read the plan first:

  1. Run: caveman learn report --json Parse the caveman.learn.v1 JSON. Show the Cave Score, its four components, and the ranked token sinks. For each sink state its class and basis. Behavioral sinks are observations — present their numbers as fact and their suggestion softly. Do not turn a behavioral finding into an imperative.

    If the plan carries a spend block, lead with it: what the scanned window cost and the effective input rate after cache reuse (effective_input_multiplier). Rules you must not break when you show money:

    • Spend is what the window COST. It is never what a fix would return.
    • Say the window it covers. Never multiply it into a month, a year, or a run rate.
    • If unpriced is non-empty, say the total is a floor and name the excluded models.
    • Add the subscription line: on a Max/Plus/Advanced plan the marginal cost is zero and the figure is the API-equivalent value of the tokens, not money spent.
    • Never call any of it verified.

Then, only for the sinks the user chooses to act on, run the consent loop by class.

Before proposing a fix, you may run: caveman learn simulate <sink_id>. Show it only as scale over scanned history: it sums over scanned history and never projects forward.

REDUCIBLE (a heavy CLAUDE.md, a never-invoked skill):

  • Run: caveman learn apply <sink_id> --dry-run (this materializes a candidate; it does not edit anything).
  • Propose a concrete diff and show before -> after tokens/turn.
  • Ask the user yes or no. On yes, apply the edit with your own file tools.
  • Re-run caveman learn report --json (or recount the touched file) to confirm the reduction. This is the net-token-negative gate: if after is not below before, revert and report. Never keep an edit that does not reduce tokens/turn.

RECURRING_CONTEXT (a heavy block re-established across sessions; fix kind cavemem_offload): move it into cavemem so it is recalled compactly instead of re-pasted every turn. The candidate carries only a LOCATOR — never the block body.

  • Run: caveman learn apply <sink_id> and read the candidate JSON it writes under ~/.caveman/candidates/. Take only the locator, the numbers, and the proposed pointer text. Do not trust any body from the candidate; there is none.
  • Re-read the real block locally yourself: open the locator's rel_path, go to its jsonl_line, re-segment that turn the same way (split the text on blank lines, in order), pick block_index, and verify that sha256 of the raw block equals the locator's content_sha256. If it does not match, the file changed since the scan — abort this item.
  • Store it: write the real block byte for byte to a temp file with your file tool, then run: caveman mem remember --stdin < and capture the returned id. Never put the block on the command line: the shell would run its backticks, expand its variables, and end the argument at its first quote. Delete the temp file once the id is back.
  • Measure the gate honestly. before = the block's tokens/turn (it loaded every turn). after = the pointer's tokens/turn plus the recall cost. Get the recall cost by running caveman mem recall "" and reading tokens_added on the hit. If after is not below before, run caveman mem forget , leave the source untouched, and stop.
  • Trim the source and write the pointer. Remove the block from its CLAUDE.md or AGENTS.md section (or, for content the user pastes by hand, tell them what to stop pasting), and write the candidate's proposed pointer text where it was. The pointer names the recall path: caveman mem recall "" for the compact form, and caveman mem recover for the byte-exact original.
  • Never make the agent dumber: before you finish, confirm that caveman mem recall "" returns a hit AND a pointer is in place. If recall returns nothing, or you did not write a pointer, REVERT (caveman mem forget and restore the source). Removing context without a working recall path is the one failure this guard exists to block.
  • Re-measure and report the confirmed reduction and the recall path.

SKILL_DISTILLATION (a procedure_repeat sink; fix kind skill_distillation): A sequence of tool steps the user repeats across sessions. Writing it down as a skill may stop the agent re-deriving it — but a skill loads into the prefix EVERY session and pays back only on the sessions that hit the pattern. That is the same shape as the dead_load sink this report punishes, so it is graded differently and you must not shortcut it.

  • Never apply this through the net-token-negative gate. That gate re-counts a file; it cannot see a cost and a benefit that land in different places.
  • Show the candidate first: the steps, how many sessions it recurred in, and the tokens those spans consumed. Say plainly that the payback is unproven.
  • If the user wants it, write the skill, then start a holdout in the same breath: caveman learn experiment start --sink <sink_id> --fix-kind skill_distillation Tell them how it works: leave it on for a stretch, then run caveman learn experiment arm <label> off and work without it for a comparable stretch. Each arm needs at least 5 sessions before any verdict exists.
  • Read the result with caveman learn experiment report <label>. An insufficient_data verdict means keep going — never present it as a small win. A regressed verdict means delete the skill; say so directly.
  • The harness compares median tokens per session. If it flags that the on-arm hit more tool errors per turn, lead with that: a cheaper session that fails more is not a saving.

MEMORY_HEALTH (memory_health::* sinks — the memory and rules doctor): audits of CLAUDE.md, CLAUDE.local.md, .claude/rules, AGENTS.md, GEMINI.md and Claude Code auto memory (MEMORY.md plus its topic files). Every item is one edit, one yes. Never delete memory content without the user's yes.

  • duplicate_rules — reducible. The same rule loads from two files every turn. Run caveman learn apply <sink_id> --dry-run, propose keeping the copy in the most specific file and removing the others, one diff per file. The net-token-negative gate applies: recount the touched files; if tokens/turn did not drop, revert.
  • memory_orphans — memory files the index never links, and index links to missing files. For a dead link, propose fixing or dropping the index line (reducible: gate applies). For an orphan file, show its first lines and ask: link it from MEMORY.md, or retire it. Linking adds index tokens — say so; that edit is outside the gate.
  • memory_truncation — MEMORY.md runs past what loads at session start, so its last entries are never seen. Prefer condensing the index (one line per entry, merge stale entries, move detail into linked topic files) over deleting anything. Show the new index and its line count against the limit before writing.
  • broken_imports — an @import points at nothing. For each one ask whether to fix the path (show the candidate file you found) or remove the import.
  • stale_references — a backticked repo path no longer exists. Behavioral: show the line and where the file likely moved; update or drop only on a yes.
  • buried_rules — a heuristic, and say so. Offer to move the listed emphatic rules nearer the top; never rewrite their wording.

LOAD_BEARING: never touch. It appears in the report only so the score stays honest.

Reporting savings (caveman learn savings):

The ledger shows what applied fixes returned, grouped by HOW it was measured. When you present it, the grouping is not decoration — it is the claim's strength:

  • deterministic_remeasure — the file we edited was re-counted. Strongest local rung.
  • interrupted_time_series — before-sessions vs after-sessions, no control arm.
  • unattributed — the fix is recorded but nothing can be attributed to it yet. Not a saving; say so.

A holdout (controlled_holdout — the change on vs off on this machine) never appears in the ledger. It comes only from caveman learn experiment report ; present it as its own result, next to the ledger, never added to it. No command produces a counterfactual_replay row yet, so never claim one.

Three rules, all binding:

  • Never sum across rungs, and never present a single blended savings headline. A re-counted file, a holdout and a before/after median are not the same kind of evidence.
  • Always read out the confounders on a row you are presenting as a win. They are standing caveats, not fine print, and they exist precisely for the good-news case.
  • Read attribution.provenance. intact means the file still carries the edit we proposed. changed_since means someone edited past it and part of the delta is not ours — say so. target_missing means the delta cannot be tied to the fix at all. not_fingerprinted means the fix predates fingerprinting, so the edit's presence is unverified. Experiments carry not_applicable: there is no single edit to check. Never present a changed_since or target_missing row as a caveman result.

A regression carries no dollar figure by design. Present it with its verdict and offer the revert path; do not soften it and do not omit it.

Binding rules:

  • Consent per edit. No "apply all" that hides the individual diffs.
  • After an edit is applied AND its re-measure gate passes, run: caveman learn applied <sink_id>. Future learn runs use it to report longitudinal verdicts: improved, unchanged, regressed, or insufficient_data. Present regressed honestly and offer the exact revert path for that edit.
  • Every edit is reversible: report exactly what you changed. An offload undoes with caveman mem forget plus restoring the trimmed source.
  • inferred only. Never present a local number as verified. Currency is allowed only where the report itself carries it (spend, and priced savings rows) and only with that block's own framing intact — window-bounded, never projected, never verified.
  • The analyzer (caveman learn) is read-only. You are the only writer, and only after a yes.
Files5
5 files · 16.2 KB

Select a file to preview

Overall Score

88/100

Grade

A

Excellent

Grades are signals, not a certification. Always review a skill yourself before use.

Safety

87

Quality

92

Clarity

89

Completeness

82

Summary

The caveman-learn skill guides an AI agent to review token consumption reports and apply user-approved edits to reduce costs: trimming heavy configuration files, offloading repeated context to a memory system (cavemem), and auditing memory files. It is a consent-gated editing workflow with strict guardrails: net-token-negative gates, never-make-the-agent-dumber safety guards, and mandatory re-measurement after each edit to verify reductions.

Detected Capabilities

read caveman learn report JSONrun caveman cli commandsfile read/write to CLAUDE.md, AGENTS.md, memory filesrun caveman mem remember/recall/forgetrun caveman learn apply and dry-runrun caveman learn applied and experiment commandspropose and confirm diffs before applyingexecute file edits on approval

Trigger Keywords

Phrases that agents use to match this skill to user intent.

lower agent token costtrim CLAUDE.mdoffload recurring contextcaveman learn reportaudit memory filesmeasure token savingsapply cost fixes

Risk Signals

INFO

File write operations on CLAUDE.md, AGENTS.md, memory files, and agent config

SKILL.md, sections REDUCIBLE, RECURRING_CONTEXT, MEMORY_HEALTH
INFO

Shell command execution: caveman learn/mem subcommands, pipelines (pipe to stdin)

SKILL.md, throughout (caveman learn report, caveman learn apply, caveman mem remember --stdin)
INFO

Outbound network: none detected; references www.apache.org only as license URL

LICENSE file

Referenced Domains

External domains referenced in skill content, detected by static analysis.

www.apache.org

Use Cases

  • Review where agent tokens are consumed
  • Trim heavy CLAUDE.md or skill definitions
  • Offload recurring context blocks to cavemem
  • Audit and deduplicate memory files and rules
  • Measure token savings from optimizations
  • Apply cost-lowering fixes with per-edit approval
  • Track improvement or regression after edits

Quality Notes

  • Excellent: skill is highly structured with clear section hierarchy (REDUCIBLE, RECURRING_CONTEXT, SKILL_DISTILLATION, MEMORY_HEALTH, LOAD_BEARING, reporting) and explicit guardrails (net-token-negative gate, never-make-dumber check, consent-per-edit, re-measurement binding rule).
  • Excellent: scope boundaries are crystal clear — which files can be edited (CLAUDE.md, AGENTS.md, .claude/rules, MEMORY.md, CLAUDE.local.md, GEMINI.md), which are read-only (caveman learn is read-only), which operations require consent (all edits), and which are forbidden (deleting memory without yes, claiming unverified savings).
  • Excellent: edge cases are well-documented — cache_efficiency is not a volume, session_outcomes is correlational, unpriced models must be named, confounders and attribution provenance must be read aloud, holdout experiments need 5+ sessions before verdict, regressed results must be presented honestly without softening.
  • Excellent: instructions include explicit revert paths (caveman mem forget <id>, restore source, re-measure, run caveman learn applied to report reversion) and verification gates (sha256 content match before offload, recall hit test before finishing, re-count after edit before accepting).
  • Excellent: the skill distinguishes three classes of sinks (reducible, recurring_context, skill_distillation, memory_health, load_bearing) and treats each with appropriate gates — deterministic re-count for files, controlled holdout for skills, honest admission of unproven payback.
  • Excellent: rules on reporting savings are binding and prevent data misuse — never sum across rungs (re-count vs before/after vs holdout), never present a single blended headline, always read confounders and attribution.provenance, never claim changed_since or target_missing rows.
  • Very good: the skill integrates with existing caveman ecosystem (caveman learn report, caveman learn apply, caveman mem, caveman learn experiment) and assumes agent has access to these tools; no installation or setup required.
  • Excellent: the skill's refusal to claim unproven savings (e.g., skill_distillation payback is unproven until holdout arm completes; unattributed rows are marked as 'not a saving'; inferred means never verified) demonstrates epistemic humility and prevents false claims.
  • Minor: the skill would benefit from a brief troubleshooting section (what to do if caveman learn report fails, if the JSON schema is unexpected, if a file changed between scan and edit, if recall returns nothing unexpectedly).
  • Minor: a worked example (a small CLAUDE.md with a repeated context block being offloaded to cavemem, showing before tokens/turn, recall cost, pointer text, and re-count) would improve clarity for first-time users, though the skill is well-written without it.
Model: claude-haiku-4-5-20251001Analyzed: Nov 18, 2026

Reviews

Add this skill to your library to leave a review.

No reviews yet

Be the first to share your experience.

Version History

  1. v2.3

    Content updated

    ✦ AIRemoves CLAUDE.md supporting file.

    2026-11-18

    LATEST
  2. v2.2

    Content updated

    ✦ AIChanges how to store blocks: now requires writing to temp file and piping stdin instead of passing block as command-line argument.

    2026-10-17

    View This Version
  3. v2.1

    Content updated

    ✦ AIAdds MEMORY_HEALTH audit sink for detecting duplicate rules, orphan files, truncation, broken imports, stale references, and buried rules in memory configuration.

    2026-10-01

    View This Version
  4. v2.0

    Contract changed: description

    ✦ AIActivation scope narrowed: removes "when the user runs caveman learn" and other explicit conditions; instructs to report savings metrics and adds procedural detail for simulating fixes, distilling…

    triggering2026-09-09

    View This Version
  5. v1.0

    2026-08-17

    View This VersionInitial version

Use JuliusBrussee/caveman-learn in your dev environment

Command Palette

Search for a command to run...