pixir-delegate

A command-line workflow for running Pixir as a headless subagent runtime from Claude Code or another compatible harness. It supports one-off workers, parallel delegation, resumable sessions, and machine-readable JSON results.

In plain words
What is it for?
Use it to run workers once, fan work out to several children, resume a session to steer it, and verify the Pixir installation and authentication.
Why use it?
It lets an agent delegate work without relying on a wrapper agent and provides session IDs and structured results for follow-up and checking.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/ranvier-technologies/pixir/pixir-delegate
Any agent
npx skills add Ranvier-Technologies/pixir --skill pixir-delegate
Clone the repo
git clone --depth 1 https://github.com/Ranvier-Technologies/pixir

Made for: Claude Code, Codex.

Per session 126 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,463 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00126 $0.04463
Opus 5 $0.00063 $0.02232
Sonnet 5 $0.00025 $0.00893
Haiku 4.5 $0.00013 $0.00446

Measured 2d ago against content hash adce9ad3cac1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

pixir-delegate 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.

The scan reads SKILL.md. This mod also ships 2 executable files (scripts/fanout.sh, scripts/steer.sh), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

.agents/skills/pixir-delegate/SKILL.md · 303 lines

How it starts

The opening of the file, as written. The whole thing — 303 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Pixir Delegate — subagents without a wrapper agent

Runtime state, hydrated by the harness at invocation. Preprocessing is this variant's contract: if the line below reads as a raw placeholder, you are on the wrong variant; use pixir-delegate-codex (explicit preflight, no hydration) instead of working around it here. pixir on PATH: !echo "$(command -v pixir || echo 'not on PATH') · v$(pixir --version 2>/dev/null || echo '?') · $(pixir doctor --json 2>/dev/null | jq -r .status 2>/dev/null || echo 'doctor unavailable')"

Know WHICH binary you are driving before delegating: if the workspace has a local build (./pixir), prefer it by explicit path over PATH — versions can differ silently. pixir doctor --json reports the local build's path under the source_install_binary check.

Pixir speaks a machine contract directly: JSON envelopes on stdout, documented exit codes, resumable session ids. The CLI is self-describing by design — discover contracts from it at runtime instead of trusting any transcript of them (including this one). Requirements: pixir on PATH, jq, authenticated (a stored pixir login credential or OPENAI_API_KEY); verify with pixir doctor --json.

Pick the right shape

Need Use
One worker, one task pixir --json --read-only - < prompt.md
N parallel workers pixir delegate --spec spec.json --json
Steer one worker over turns pixir --json resume <sid> "next task"
Fan out, then steer or retry one child delegate, then resume <child_session_id>

Do NOT spawn N independent one-shot processes for fan-out — each boots its own VM. delegate exists so N children share one runtime. Prefer resume over fresh sessions for follow-ups: a resumed turn reuses the session's history and prompt-cache key, but do NOT budget on guaranteed cache hits (a blind-run measured 0% cached on a resume minutes after the fan-out; hits depend on provider cache state). Session ids look global but resume resolves them against $PWD/.pixir/sessions/ only — run it from the workspace root, not from a scratchpad.

Read the full file on GitHub · 303 lines

Files

What ships with it

4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

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.

  1. 2d ago First seen · 303 lines · 126 tokens per session scan A adce9ad3cac1

Subscribe to this mod's changes

pixir-delegate is a skill published in the GitHub repository Ranvier-Technologies/pixir (23 stars, last pushed 5d ago), licensed MIT. It adds 126 tokens to every session and 4,463 once invoked, about $0.0006 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 2026-08-30.

Related

Other skills, from other repositories

cross-linker

Scan the Obsidian wiki and automatically discover missing cross-references between pages. Use this skill when the user says "link my pages", "find missing links", "cross-reference", "connect my wiki", "add wikilinks", "what pages should be linked", or after any large ingestion to ensure new pages are woven into the…

saski/arnesto · 137 tokens

free-agent-execution

Run an explicitly requested, bounded, non-sensitive coding task through the local free-worker adapter using a verified free OpenCode model. Use only when the user asks to execute a free worker or free agent and accepts the external-model safety boundary.

saski/arnesto · 52 tokens

google-adk-agent-patterns

Define, compose, and wire Google ADK agents with function tools and multi-agent orchestration. Use when writing agent definitions, adding Python function tools, building pipelines with SequentialAgent/LoopAgent/ParallelAgent, or designing router/delegation patterns. Trigger on: adk agent, LlmAgent, SequentialAgent…

saski/arnesto · 92 tokens

corporate-aws-cli

Use when using the AWS CLI against a corporate AWS account after SAML/OIDC federation (e.g. saml2aws, aws sso login), when STS fails with invalid XML or empty responses, when selecting profiles or regions for Terraform or CI/CD stacks, or when validating RDS engine versions with describe-orderable-db-instance-options…

saski/arnesto · 81 tokens

google-adk-setup

Bootstrap a Google ADK (Agent Development Kit) project from zero to a running local dev UI. Use when setting up a new ADK agent project, configuring Vertex AI credentials, scaffolding the agent folder, or troubleshooting ADK environment issues. Trigger on: adk setup, adk init, google-adk, vertex ai agent, uv run adk…

saski/arnesto · 83 tokens

align

Presents a proposed approach in progressive confirmable chunks with recommended decisions and alternatives. Use when aligning on a design, plan, or technical approach before implementation.

saski/arnesto · 33 tokens