caveman
claudeTalk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Records every prompt you send to Claude Code, scores it against a rubric, and builds a personal prompting guide with before/after examples.
Plugin prompt-journal from marketplace prompt-journal-marketplace in waqar40/prompt-improvement-framework. Commands read from the repo's manifest — 1 month ago.
This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.
artifact-reviewerUse to REVIEW existing Claude Code assets — a skill, subagent, hook, slash command, memory rule, or script (one asset or a whole .claude/ tree) — against the same quality gate we build to. Runs the deterministic frontmatter validator, then audits anatomy, the 7 rubrics (correctness/latency/cost/secu
asset-architectUse to decide what kind of Claude Code asset a need should become — a skill, subagent, hook, slash command, memory rule, script, or plugin — and where it should live, then scaffold it per Anthropic best practices. A multi-source grounding + deep-research consumer: before drafting, it reads the targe
asset-fixerUse to APPLY the mechanical, fully-specified fixes from an artifact-reviewer findings report to an existing Claude Code asset (skill/agent/hook/command/rule/script) — never to redesign or author new content. Takes the reviewer's findings (or re-runs the review itself), applies only fixes a reviewer
asset-suggesterUse after prompts have been scored (by prompt-critic, via the prompt-journal pipeline) to detect recurring work that should become a reusable Claude Code asset — a skill, subagent, hook, slash command, rule, or script — and record deduped candidates in suggestions/<user>.json. Reads the score store
catalogUse to print the prompt-journal framework's capability guide — a catalog of every command, skill, and agent in this repo with its usage, inputs, and outcomes. Triggers include "help", "list commands", "list skills", "what can this repo do", "what commands are available", "show me the skills", and th
configureUse to verify (or repair) the prompt-journal recorder after installing this plugin — the /configure command. As an installed plugin all three recorder hooks (UserPromptSubmit, PostToolUse, Stop) wire themselves automatically via hooks/hooks.json; this skill's job is to make sure the rest of the setu
progress-coachUse as a step inside the /analyse pipeline (after scoring, before prompt-example-curator) to turn a user's score history into an adaptive, personalized coaching plan — identifies the current strongest/weakest rubric dimensions, tracks pace of improvement between /analyse runs, decides the single nex
prompt-criticUse when a user asks to review, critique, score, rate, evaluate, or improve a prompt written for an LLM — either a one-shot user prompt or a reusable system prompt. Scores the prompt against a two-layer rubric (design + evaluability) distilled from Anthropic's and OpenAI's guidance, localizes each g
prompt-example-curatorUse after prompt-critic has scored one or more prompts, when the user wants to classify prompts as bad / good / excellent and turn the best and worst into worked examples in their personal prompting guide. Reads prompt-critic JSON output (or a score store), assigns each prompt a band, selects the mo
prompt-journalUse to run the end-to-end prompt-journal pipeline over recorded logs — by default the WHOLE journal (every log across every project/branch), or narrowed to one project, branch, or file. Scores every prompt with prompt-critic (treating the entries under one branch as one session/chain), writes a per-
test-frameworkUse to end-to-end TEST this prompt-journal framework — the /test play. Runs the deterministic script harness, then drives every key skill and command (prompt-critic, prompt-journal/analyse with all its selectors, progress-coach, prompt-example-curator, asset-suggester, asset-architect, artifact-revi
version-control-shortcutUse when the user gives a vague git instruction like "commit this", "push it", or "version-control it" instead of naming the exact action, file scope, or message. Resolves the intended git operation from `git status`/`git diff` and the conversation, guards against staging this repo's gitignored prom
/prompt-journal:analyseAnalyse prompts — by default the WHOLE journal (all projects/branches), or narrowed to one --project/--branch/file. Scores each prompt, writes per-file reviews, and compiles the overall guide + machine-readable asset suggestions.
/prompt-journal:catalogList everything in this repo — every command, skill, and agent with its usage, inputs, and outcomes.
/prompt-journal:configureVerify/repair the prompt-journal recorder after install — detects your OS, confirms the plugin's UserPromptSubmit hook is working, auto-fixes issues (dirs, deps), self-tests, and guides you only if something needs your hands.
/prompt-journal:fix-assetApply the mechanical, fully-specified fixes from an artifact-reviewer findings report to an existing asset (dangling references, invalid frontmatter keys, stale table rows). Never authors new content — anything needing judgment is skipped and routed to /scaffold-asset.
/prompt-journal:prompt-reviewAlias of /analyse — run the prompt-journal pipeline over a log or directory (score prompts, update the store + guide, refresh asset suggestions).
/prompt-journal:review-assetReview existing Claude Code assets (skill/agent/hook/command/rule/script — one or a whole .claude/ tree) against the shared quality gate: frontmatter, anatomy, the 7 rubrics, non-destructive permissions, model tier, a shipped (on-disk) verification, and plan-first/Responsible AI. Read-only — reports
/prompt-journal:scaffold-assetDecide the right asset type + placement for a need and scaffold it (skill/subagent/hook/command/rule/script/plugin), grounded + deep-researched (code, topic, edge cases, Responsible AI, memory files). Asks rather than assumes, presents a written plan + draft before writing, and generates a real vali
/prompt-journal:testEnd-to-end test the framework — runs the script harness, then drives every key skill/command over fixtures in an isolated _selftest sandbox and reports PASS/FAIL. Touches no real data.
bash "${CLAUDE_PLUGIN_ROOT}/scripts/record-prompt.sh"
bash "${CLAUDE_PLUGIN_ROOT}/scripts/record-tool-use.sh"
bash "${CLAUDE_PLUGIN_ROOT}/scripts/record-turn-end.sh"
Talk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
5 skills · hooks
Cross-session memory and token savings for coding agents.
6 skills · 1 agent · 1 MCP server · hooks
Package any website or local web build into a lightweight desktop app with Pake CLI.
1 skill
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.