Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/mduffy37/claudeworks/create-profilenpx skills add Mduffy37/claudeworks --skill create-profilegit clone --depth 1 https://github.com/Mduffy37/claudeworksWhat 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 | $0.00042 | $0.16157 |
| Opus 5 | $0.00021 | $0.08079 |
| Sonnet 5 | $0.00008 | $0.03231 |
| Haiku 4.5 | $0.00004 | $0.01616 |
Grade B, and why
create-profile scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
!`cat "${CLAUDE_PLUGIN_ROOT:-$HOME/.claude/plugins/marketplaces/claudeworks/plugins/profiles-manager}/data/workflow-shapes.json"` How it starts
The opening of the file, as written. The whole thing — 678 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are creating a new Claude Code profile. A profile is a named preset that controls which plugins, skills, agents, MCP servers, and settings load per Claude Code session. Your job is not to match keywords to plugins — it is to produce a workflow-shaped profile that covers the natural stages of the user's work with tools that fit their specific tech stack and constraints, and then to turn that draft into a finished profile through a short conversation with the user.
This skill runs as a layered flow:
- Step 0 — Description. Capture a one-line profile description from the user's initial message (or by asking directly). Everything downstream grounds against this intent.
- Layer 0 — Project inference. If launched inside a code project, infer the tech context before asking questions.
- Layer 1 — Clarification. Adaptive 0–3 questions depending on what Layer 0 already knows.
- Layer 2 — Retrieval. For each stage of the chosen workflow shape(s), pull candidate plugins from the full curated marketplace using stage keywords × tech context.
- Layer 3 — Composition. Assemble a draft plugin list, then self-critique it against a gap / redundancy / coherence checklist.
- Layer 4 — Interactive presentation. Discuss the picks with the user, co-design an optional
/workflowcommand, collect final settings, then writeprofiles.json. - Step 7.5 — Plugin install. For any final picks not yet installed on the machine, offer to install them before the write so the profile launches cleanly.
You never read the marketplace index in full. You use grep and jq to filter, and only ever see the subset of plugins that match the user's actual intent. And you never draft the /workflow body unilaterally — that is always co-designed with the user in Step 7d.
Step 0 — Capture the profile description
Before running any inference or retrieval, lock a one-line description for this profile. This text appears next to the profile name in the ClaudeWorks app's sidebar and anchors every downstream decision — shape matching, keyword derivation, presentation rationale all reference it.
Read the user's initial message. If it already contains enough signal to draft a plausible description (e.g. "I'm building frontend features on this app — React + TS, Tailwind, Playwright" becomes "Frontend feature development on a React/TS/Tailwind solo project"), draft one and present it for a single-beat confirmation:
"Before I run project inference, let me lock a one-line description for this profile (it'll show up next to the profile name in the ClaudeWorks app). How about: 'Frontend feature development on a React/TS desktop app'? Sound right, or do you want a different wording?"
Accept the user's confirmation or revision. Store the final description as the profile_description you'll reference in Step 3 (shape matching), Step 7a (presentation header), and Step 8 (as P_DESC when writing to profiles.json).
If the user's initial message is too thin to draft from (e.g. "I want a profile" or "make me a profile"), ask directly:
"What's this profile for, in one line? It'll show up in the profile list — something like 'Frontend features on my React app' or 'PR review for the backend team'. Be as specific as you want."
Do not move on from Step 0 until the description is locked. A vague description produces a vague profile; you can recover from a weak Layer 0 signal bundle, but you cannot recover from a user whose intent you never understood.
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.
- 2d ago First seen · 678 lines · 42 tokens per session scan B a889102a2319
create-profile is a skill published in the GitHub repository Mduffy37/claudeworks (11 stars, last pushed 3mo ago), licensed MIT. It adds 42 tokens to every session and 16,157 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
todos
This chat has a shared, live TODO plan — your tasks for the conversation, which the user also edits. Read this skill and reach for the todo tools whenever a request takes more than a couple of steps. It covers the plan model (group = task, items = its steps; loose items are the user's lane), how to work it: propose…
writing-workflow-skills
Use when adding a new workflow skill to pi-thinkrail-workflow, changing an existing workflow skill's role, trigger, handoff, or structure, or checking a workflow skill against the workflow system's rules. Not for authoring general-purpose skills outside this package.
brainstorming
Use this BEFORE any creative or feature work: building a new feature, adding functionality, changing behavior, or making a nontrivial design decision. Turns the user's request into a validated design — recorded as a spec-graph task-spec — before any implementation. Do not skip this because a change looks small.
shipping-a-pr
Use when finished work needs to ship as a pull request, or when the ask is about a PR — creating one, bringing it up to date, adding screenshots, watching its checks, or addressing its review comments. Not for reviewing a PR you are not shipping.
writing-specs
Use when a workflow step drafts or revises a spec artifact — a goal-and-requirements, an architecture, or a module SPEC — or when a workflow skill names it at such a step. The shared quality bar for specs — not a workflow, nothing to execute.
clarify
Adaptive requirements clarification with auto-depth routing. Shallow (Q&A) for simple tasks, Deep (exploration + DRAFT + PLAN) for complex ones. Escalates automatically when ambiguity persists.