internal/seed/sidecar/skills/create-from-project/SKILL.md
# Create Identity from Project Context Create an ethos identity by fine-tuning starter content from your existing project files. 90% batteries-included defaults, 10% LLM-inferred personalization from what you already have. ## When to Use Run this when you've just installed ethos and want to create your first identity without starting from scratch. The skill reads your project context (CLAUDE.md, .claude/agents/, git config) and proposes identity attributes that match your existing preferences
npx skillsauth add punt-labs/ethos internal/seed/sidecar/skills/create-from-projectInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Create an ethos identity by fine-tuning starter content from your existing project files. 90% batteries-included defaults, 10% LLM-inferred personalization from what you already have.
Run this when you've just installed ethos and want to create your first identity without starting from scratch. The skill reads your project context (CLAUDE.md, .claude/agents/, git config) and proposes identity attributes that match your existing preferences.
Read the following files if they exist:
user.name and user.email via
git config user.name and git config user.emailIf none of these exist, fall back to the interactive wizard:
ethos create.
From the gathered context, propose:
user.nameuser.emailconcise-quantifiedprecise-writermeasured-proseconcise-quantifiedprincipal-engineer for most developers.engineering
Python project → include engineering
Multiple languages → include engineeringShow the proposed identity to the user:
Based on your project context, I'd create this identity:
Name: Jim Freeman
Handle: jfreeman
Email: [email protected]
Kind: human
Writing style: concise-quantified
(inferred from: "Keep sentences under 30 words"
in CLAUDE.md)
Personality: principal-engineer
(inferred from: "You are a principal engineer"
in CLAUDE.md)
Talents: engineering
Accept this identity, or adjust any field?
Use AskUserQuestion to confirm. The user can accept as-is or adjust individual fields.
Call ethos create -f <temp-yaml> with the confirmed attributes.
Or use the MCP identity tool with method: create.
Report what was created and suggest next steps:
Created identity: jfreeman (human)
Writing style: concise-quantified
Personality: principal-engineer
Next steps:
- Create agent personas: ethos create (with kind: agent)
- Configure your repo: add agent: <handle> to .punt-labs/ethos.yaml
- Restart Claude Code to activate
ethos create or
ethos import --from soulspec for agentsdevelopment
# Mission Scaffold a Phase 3 mission contract, register it in the store, and spawn the worker. This skill turns a conversation about "who does what" into a typed, enforced delegation — the write set is admitted, the evaluator is frozen, rounds are bounded, the event log records every transition. ## Who invokes this Leaders only. Sub-agents cannot spawn other sub-agents (Claude Code constraint), so this skill is a no-op inside a delegation. If you are reading this from inside an `Agent()` call
tools
# Baseline Operations Operational discipline for sub-agents. This skill replaces the operational guidance lost when Claude Code's default system prompt is replaced by an agent definition. ## Tool Usage Use dedicated tools instead of shell equivalents: - **Read** files, never `cat`, `head`, `tail`, or `sed` - **Grep** for content search, never `grep` or `rg` via Bash - **Glob** for file discovery, never `find` or `ls` via Bash - **Edit** for modifications, never `sed` or `awk` via Bash - **Wr
data-ai
Example TaskFlow authoring pattern for inbox triage. Use when messages need different treatment based on intent, with some routes notifying immediately, some waiting on outside answers, and others rolling into a later summary.
data-ai
Example TaskFlow authoring pattern for inbox triage. Use when messages need different treatment based on intent, with some routes notifying immediately, some waiting on outside answers, and others rolling into a later summary.