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/ohswedd/praxis/bootstrapnpx skills add Ohswedd/praxis --skill bootstrapgit clone --depth 1 https://github.com/Ohswedd/praxisWhat 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 | $0.00186 | $0.02745 |
| Opus 5 | $0.00093 | $0.01373 |
| Sonnet 5 | $0.00037 | $0.00549 |
| Haiku 4.5 | $0.00019 | $0.00275 |
Grade A, and why
bootstrap scanned grade A with 0 findings 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 181 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bootstrap
Bring a repository to a top-tier Claude Code setup, whatever state it starts in. This is language- and framework-agnostic: infer the stack from what is actually present; do not assume a fixed set of ecosystems.
Step 0a: Automatic invocation (the normal case)
This skill is not usually asked for. The SessionStart audit and the prompt router
both instruct you to run it whenever repo_state is not managed, because a repo
with no operating brief, no guardrails and no living knowledge is the state praxis
exists to prevent, and a recommendation to fix it was routinely stepped past.
When it fires automatically, it is a step, not a conversation:
- Write what does not exist, without asking. Creating a file the repo lacks is additive and reversible, and the user asked for the work, not for a setup interview.
- Then carry straight on to the user's actual request, in the same turn. Report the setup in one or two lines under a "Setup" heading in your reply. Bootstrap that hijacks the first prompt is worse than bootstrap that never ran.
- Stop and ask for exactly one thing: reconciling a
CLAUDE.mdthat praxis did not author (Step 4). That merge can drop a still-valid instruction, so it gets the verifier and the user's eyes. Nothing else here is lossy. - If the user has turned it off (
/praxis:config bootstrap off,bootstrap.auto = false,PRAXIS_BOOTSTRAP=off), do not run it and do not argue about it.
Step 0b: Resolve the workspace mode (before writing anything)
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/config.py" status names the mode and
where it came from. It decides every path below:
owner |
contributor |
|
|---|---|---|
| Operating brief | CLAUDE.md (+ nested) |
CLAUDE.local.md at the root only |
| Settings | .claude/settings.json |
.claude/settings.local.json |
| praxis config | .praxis.toml (committed) |
.claude/.praxis/praxis.toml (local) |
/docs, CHANGELOG.md, docs/adr/ |
created and maintained | updated only if the repo already has them; otherwise written under .claude/.praxis/knowledge/ |
| Ignore rules | add praxis's paths to .gitignore |
never touch .gitignore; praxis maintains $GIT_COMMON_DIR/info/exclude itself |
| Capabilities (Step 7) | .claude/ commands, skills, agents; .mcp.json |
.claude/.praxis/capabilities/; MCP in .claude/settings.local.json |
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 · 181 lines · 186 tokens per session scan A 2379e3cb4bf8
bootstrap is a skill published in the GitHub repository Ohswedd/praxis (1 stars, last pushed 1mo ago), licensed MIT. It adds 186 tokens to every session and 2,745 once invoked, about $0.0009 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
github-workflow-standards
Core standards for all GitHub workflow agents. Covers authentication, smart defaults, repository discovery, dual MD+HTML output, screen-reader-compliant HTML accessibility standards, safety rules, progress announcements, parallel execution, and output quality. Apply when building any GitHub workflow agent - issues…
Developer Hub
Your intelligent developer command center -- start here for any Python, wxPython, desktop app, NVDA addon, accessibility tool building, desktop accessibility, or general software engineering task. Routes to specialist agents across the developer, web, and document accessibility teams.
agent-memory-provenance
Design attributable, privacy-aware memory for AI agents. Use when an agent needs session context, durable memory, retrieval, summaries, source citations, client isolation, compaction, retention rules, or an append-only decision and evidence record.
agent-orchestration-architecture
Design reliable AI-agent and multi-agent systems. Use when deciding whether a workflow needs one agent, tools, specialist agents, manager control, handoffs, durable execution, human approval, model routing, retries, or an operating boundary for an AI workflow.
using-humane
Route a design task to the right humane skill, and after any humane skill finishes, name the next one from the declared handoff graph instead of guessing. Reads the handoffs: / accepts: / orchestrates: keys in each skill's frontmatter, so the route is data, not recollection. Use at the start of design work when it is…
agentic-app-risk-review
Defensive review of an LLM/agent application for OWASP-LLM-style risks — prompt injection, unsafe tool calls, excessive agency, and PII/secret leakage. Reads your code, flags issues with file:line. Pure Claude Code, no signup.