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 skills add thienhoc/dwi-by-thienhoc --skill dwi-arcgit clone --depth 1 https://github.com/thienhoc/dwi-by-thienhocWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/thienhoc/dwi-by-thienhoc/dwi-arc)<a href="https://agentmods.dev/skills/thienhoc/dwi-by-thienhoc/dwi-arc"><img src="https://agentmods.dev/badge/skills/thienhoc/dwi-by-thienhoc/dwi-arc/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/thienhoc/dwi-by-thienhoc/dwi-arc"><img src="https://agentmods.dev/badge/skills/thienhoc/dwi-by-thienhoc/dwi-arc.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00058 | $0.00569 |
| Opus 5 | $0.00029 | $0.00284 |
| Sonnet 5 | $0.00012 | $0.00114 |
| Haiku 4.5 | $0.00006 | $0.00057 |
Grade A, and why
dwi-arc 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 11d 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 — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Dwi Arc
Use a layered multi-agent structure only when it lowers risk or time for real work.
Entry check
Do not create an arc for a task one agent can complete clearly and safely.
Use Arc when at least one is true:
- independent review materially improves confidence;
- several workstreams are genuinely disjoint;
- a specialist context is required;
- integration needs a separate owner;
- the task has an explicit multi-stage acceptance gate.
Roles
Root
Owns the user outcome, authority, cell boundaries, integration decision, and final truth.
Work cell
Owns one bounded question or write scope. A cell does not expand its own authority.
Independent gate
Reviews evidence or behavior without sharing the writer's assumptions. It reports findings; Root decides acceptance.
Stage pattern
Use only the stages the task needs:
| Stage | Purpose |
|---|---|
| W1 | Produce one bounded analysis or implementation |
| W2 | Integrate compatible outputs into the shared result |
| W3 | Independently audit behavior, risk, or evidence |
| Root gate | Accept, return, narrow, or stop |
Do not create empty stages for symmetry.
One-writer discipline
- Each mutable scope has one writer.
- Parallel reads are allowed.
- Parallel writes require explicitly disjoint files or systems.
- Shared capacities such as a branch, release, schema, or deployment remain serialized.
- If two active writers overlap, stop the conflicting scope until ownership is clear.
Cell packet
Give each cell:
Outcome:
Read scope:
Write scope, or read-only:
Non-goals:
Acceptance evidence:
Return format:
Do not send unrelated private context.
Root protocol
- Resolve the live workspace and user authority.
- Choose the fewest useful cells.
- Assign disjoint scopes and one writer each.
- Run parallel work only when independence is real.
- Integrate through Root, not peer-to-peer mutation.
- Use an independent gate for material claims.
- Report accepted work, rejected work, failures, and unknowns separately.
What ships with it
1 file 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.
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.
- 11d ago First seen · 89 lines · 58 tokens per session scan A fd45ab58e960
dwi-arc is a skill published in the GitHub repository thienhoc/dwi-by-thienhoc (3 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 58 tokens to every session and 569 once invoked, about $0.0003 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
awsl
Run Claude Code JavaScript Workflows through the awsl compatibility runtime. Use when an agent's current task or loaded Skill requires dispatching a Claude Code Workflow but the host cannot execute that Workflow natively, or when awsl workflow inspection, durable run state, resume, or provider diagnostics are needed.
grill-with-docs-lite
Groom a fuzzy request and the smallest relevant source set into a bounded PRE-SPEC ASSEMBLY packet for mini-spec, separating evidence from decisions, assumptions, and blockers before contract language is written.
handoff
Compress project context into HANDOFF.md with workflow state, analysis checkpoint, next gate, verification, current hypothesis, freshness anchors, optional gotcha references, and a resume packet for the next agent session.
tool-noise-guard
Passive guardrail that compacts repetitive tool-result envelopes, preserves decision-relevant anchors and deltas, and suppresses redundant re-fetches without pretending prior context disappeared.
verify-contract
Record clear evidence that a task works, including commands, results, remaining risks, scope adherence, and whether the implementation stayed under the spec ceiling.
analyze-mini
Risk-triggered consistency checkpoint that detects when task artifacts, assumptions, or verification no longer support a safe build decision.