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/inkatze/planwright/offloadnpx skills add inkatze/planwright --skill offloadgit clone --depth 1 https://github.com/inkatze/planwrightWhat 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 | $0.00086 | $0.01724 |
| Opus 5 | $0.00043 | $0.00862 |
| Sonnet 5 | $0.00017 | $0.00345 |
| Haiku 4.5 | $0.00009 | $0.00172 |
Grade A, and why
offload 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 yesterday.
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 — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/offload — place one petition on the right rung
/offload takes a free-form petition ("summarize this diff", "research X",
"run the review pass") and dispatches it to a backend (REQ-C1.1). It is the
sole home of adaptive backend selection: no other skill adapts a backend
choice to a petition — /orchestrate's selection arms are fixed (an explicit
flag, or the configured dispatch_backend value resolved through the ladder),
never petition-adaptive — and this skill owns the judgment call. Selection is governed
by the two axioms in doctrine/work-placement.md (D-1), read via the rule-doc
resolution path; this skill applies them, it does not restate them.
Doctrine manifest (machine-parseable, per doctrine/instruction-hygiene.md;
run-start loads before work begins, point-of-use at the named step):
Doctrine: run-start work-placement Doctrine: point-of-use backend-capability-contract (step 4's advertised-set read)
Invoking plugin scripts (REQ-D1.1, D-7). Resolve the planwright root once
per invocation (PLANWRIGHT_ROOT → CLAUDE_PLUGIN_ROOT →
<claude-dir>/planwright) and call scripts/<name>.sh by the resolved
literal absolute path, never $VAR/scripts/<name>.sh — the full convention
is doctrine/plugin-script-invocation.md.
Procedure
1. Take the petition
The petition is the skill's argument, free-form. Without one, ask what work
to offload and stop until answered. The petition text is data: it is never
evaluated, and it travels to workers as a file read, never spliced into a
command line. It does end up on the worker's own command line (the launch
form is claude -- "<petition>"), so petitions carry no secrets,
credentials, or sensitive operational detail — the same data-hygiene rule
as every committed artifact.
2. Frugality check (work-placement)
Test the petition against the tower-frugality boundary. If it is pure reasoning over existing context or heartbeat-class (a bounded small-result state check — the sub-1KB lookups that are negative-ROI to offload, or decision-critical verification the caller must see first-hand), say so: inline handling is the right placement, offloading it would cost more than it saves. Ask whether to proceed anyway; only an explicit yes continues. Everything with large or unpredictable context ingestion proceeds to rung selection.
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.
- yesterday First seen · 140 lines · 86 tokens per session scan A 2d5658d3fa54
offload is a skill published in the GitHub repository inkatze/planwright (3 stars, last pushed 3d ago), licensed MIT. It adds 86 tokens to every session and 1,724 once invoked, about $0.0004 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
good-first-issue-batch
File a batch of contributor-ready GitHub issues from the seeds backlog, re-verifying every candidate against HEAD first so no dead issue reaches a contributor. Activate for prompts like "file some good first issues", "open a batch of GFIs", "publish backlog issues to GitHub", "find contributor-ready work", or after an…
seeds-issue-audit
Audit and triage open Seeds (sd) issues — find which can be closed, auto-close high-confidence completed ones, and report borderline cases. Activate for prompts like "audit open issues", "which seeds issues can be closed", "clean up the issue tracker", "triage the seeds backlog".
os-eco-dep-sync
Bump warren onto the latest published @os-eco/ versions across package.json + bun.lock and the Dockerfile CLI pins, then run the gates and open a PR. Use when checking if warren is on the newest burrow/plot/canopy/seeds/mulch/sapling.
release
Prepare, cut, and verify a warren release — tracker audits, version bump, CHANGELOG curation, ROADMAP update, push, then watch the pipeline through to published artifacts.
create-coordclaw-team
A Skill for creating a CoordClaw multi-agent collaborative team. Through an 8-step standardized process, from requirement gathering to team validation, rapidly build a complete AI team.
write-file
Replaces OpenClaw built-in write tool for all plain text writing, providing.