cavecrew

Name: cavecrew
Author: JuliusBrussee

208kinstalls5.2kforks

How to install cavecrew?

juliusbrussee/caveman/cavecrew

$npx -y skills add JuliusBrussee/caveman --skill cavecrew

Installs into the current project.

Summary

Decision guide for delegating to caveman-style subagents. Tells the main thread WHEN to spawn cavecrew-investigator (locate code), cavecrew-builder (1-2 file edit), or cavecrew-reviewer (diff review) instead of doing the work inline or using vanilla Explore. Subagent outp

Files · 1

View on GitHub

SKILL.md

1	Cavecrew = three subagent presets that emit caveman output. Same job as Anthropic defaults (`Explore`, edit-style agents, reviewer); difference is the tool-result they return is compressed, so main context shrinks per delegation.
2
3	## When to use cavecrew vs alternatives
4
5	\| Task \| Use \|
6	\|---\|---\|
7	\| "Where is X defined / what calls Y / list uses of Z" \| `cavecrew-investigator` \|
8	\| Same but you also want suggestions/architecture commentary \| `Explore` (vanilla) \|
9	\| Surgical edit, ≤2 files, scope obvious \| `cavecrew-builder` \|
10	\| New feature / 3+ files / cross-cutting refactor \| Main thread or `feature-dev:code-architect` \|
11	\| Review diff, branch, or file for bugs \| `cavecrew-reviewer` \|
12	\| Deep code review with rationale + alternatives \| `Code Reviewer` (vanilla) \|
13	\| One-line answer you already know \| Main thread, no subagent \|
14
15	Rule of thumb: if you'd want the subagent's output in 1/3 the tokens, pick cavecrew. If you'd want prose, pick vanilla.
16
17	## Why this exists (the real win)
18
19	Subagent tool results get injected into main context verbatim. A vanilla `Explore` that returns 2k tokens of prose costs 2k tokens of main-context budget every time. The same finding from `cavecrew-investigator` returns ~700 tokens. Across 20 delegations in one session that's the difference between context exhaustion and finishing the task.
20
21	## Output contracts
22
23	What main thread can rely on per agent:
24
25	`cavecrew-investigator`
26	```
27	<Header>:
28	- path:line — `symbol` — short note
29	totals: <counts>.
30	```
31	Or `No match.` Always file-path-first, line-number-attached, backticked symbols. Safe to grep with `path:\d+`.
32
33	`cavecrew-builder`
34	```
35	<path:line-range> — <change ≤10 words>.
36	verified: <re-read OK \| mismatch @ path:line>.
37	```
38	Or one of: `too-big.` / `needs-confirm.` / `ambiguous.` / `regressed.` (terminal first token).
39
40	`cavecrew-reviewer`
41	```
42	path:line: <emoji> <severity>: <problem>. <fix>.
43	totals: N🔴 N🟡 N🔵 N❓
44	```
45	Or `No issues.` Findings sorted file → line ascending.
46
47	## Chaining patterns
48
49	Locate → fix → verify (most common):
50	1. `cavecrew-investigator` returns site list.
51	2. Main thread picks 1-2 sites, hands paths to `cavecrew-builder`.
52	3. `cavecrew-reviewer` audits the diff.
53
54	Parallel scout (when investigation is broad):
55	Spawn 2-3 `cavecrew-investigator` calls in one message (different angles: defs vs callers vs tests). Aggregate in main thread.
56
57	Single-shot edit (when site is already known):
58	Skip investigator. Hand exact path:line to `cavecrew-builder` directly.
59
60	## What NOT to do
61
62	- Don't use `cavecrew-builder` when you don't already know the file. Spawn investigator first or main thread will eat tokens passing context.
63	- Don't chain `cavecrew-investigator → cavecrew-builder` for a 5-file refactor. Builder will return `too-big.` and you'll have wasted a turn.
64	- Don't ask `cavecrew-reviewer` for "general feedback" — it returns findings only, no architecture opinions. Use `Code Reviewer` for that.
65	- Don't expect prose. Cavecrew output is structured, sometimes terse to the point of cryptic. If a human will read it directly, paraphrase.
66
67	## Auto-clarity (inherited)
68
69	Subagents drop caveman → normal English for security warnings, irreversible-action confirmations, and any output where fragment ambiguity could be misread. Resume caveman after.

People who liked this also viewed

6 picks

Type

cavecrew

208kinstalls5.2kforks

View on GitHub

How to install cavecrew?

juliusbrussee/caveman/cavecrew

$npx -y skills add JuliusBrussee/caveman --skill cavecrew

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	Cavecrew = three subagent presets that emit caveman output. Same job as Anthropic defaults (`Explore`, edit-style agents, reviewer); difference is the tool-result they return is compressed, so main context shrinks per delegation.
2
3	## When to use cavecrew vs alternatives
4
5	\| Task \| Use \|
6	\|---\|---\|
7	\| "Where is X defined / what calls Y / list uses of Z" \| `cavecrew-investigator` \|
8	\| Same but you also want suggestions/architecture commentary \| `Explore` (vanilla) \|
9	\| Surgical edit, ≤2 files, scope obvious \| `cavecrew-builder` \|
10	\| New feature / 3+ files / cross-cutting refactor \| Main thread or `feature-dev:code-architect` \|
11	\| Review diff, branch, or file for bugs \| `cavecrew-reviewer` \|
12	\| Deep code review with rationale + alternatives \| `Code Reviewer` (vanilla) \|
13	\| One-line answer you already know \| Main thread, no subagent \|
14
15	Rule of thumb: if you'd want the subagent's output in 1/3 the tokens, pick cavecrew. If you'd want prose, pick vanilla.
16
17	## Why this exists (the real win)
18
19	Subagent tool results get injected into main context verbatim. A vanilla `Explore` that returns 2k tokens of prose costs 2k tokens of main-context budget every time. The same finding from `cavecrew-investigator` returns ~700 tokens. Across 20 delegations in one session that's the difference between context exhaustion and finishing the task.
20
21	## Output contracts
22
23	What main thread can rely on per agent:
24
25	`cavecrew-investigator`
26	```
27	<Header>:
28	- path:line — `symbol` — short note
29	totals: <counts>.
30	```
31	Or `No match.` Always file-path-first, line-number-attached, backticked symbols. Safe to grep with `path:\d+`.
32
33	`cavecrew-builder`
34	```
35	<path:line-range> — <change ≤10 words>.
36	verified: <re-read OK \| mismatch @ path:line>.
37	```
38	Or one of: `too-big.` / `needs-confirm.` / `ambiguous.` / `regressed.` (terminal first token).
39
40	`cavecrew-reviewer`
41	```
42	path:line: <emoji> <severity>: <problem>. <fix>.
43	totals: N🔴 N🟡 N🔵 N❓
44	```
45	Or `No issues.` Findings sorted file → line ascending.
46
47	## Chaining patterns
48
49	Locate → fix → verify (most common):
50	1. `cavecrew-investigator` returns site list.
51	2. Main thread picks 1-2 sites, hands paths to `cavecrew-builder`.
52	3. `cavecrew-reviewer` audits the diff.
53
54	Parallel scout (when investigation is broad):
55	Spawn 2-3 `cavecrew-investigator` calls in one message (different angles: defs vs callers vs tests). Aggregate in main thread.
56
57	Single-shot edit (when site is already known):
58	Skip investigator. Hand exact path:line to `cavecrew-builder` directly.
59
60	## What NOT to do
61
62	- Don't use `cavecrew-builder` when you don't already know the file. Spawn investigator first or main thread will eat tokens passing context.
63	- Don't chain `cavecrew-investigator → cavecrew-builder` for a 5-file refactor. Builder will return `too-big.` and you'll have wasted a turn.
64	- Don't ask `cavecrew-reviewer` for "general feedback" — it returns findings only, no architecture opinions. Use `Code Reviewer` for that.
65	- Don't expect prose. Cavecrew output is structured, sometimes terse to the point of cryptic. If a human will read it directly, paraphrase.
66
67	## Auto-clarity (inherited)
68
69	Subagents drop caveman → normal English for security warnings, irreversible-action confirmations, and any output where fragment ambiguity could be misread. Resume caveman after.

People who liked this also viewed

6 picks

Type