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/rhettcarillo/claude-code-direct-skill/directnpx skills add rhettcarillo/claude-code-direct-skill --skill directgit clone --depth 1 https://github.com/rhettcarillo/claude-code-direct-skillWrote 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/rhettcarillo/claude-code-direct-skill/direct)<a href="https://agentmods.dev/skills/rhettcarillo/claude-code-direct-skill/direct"><img src="https://agentmods.dev/badge/skills/rhettcarillo/claude-code-direct-skill/direct.svg" alt="Measured on agentmods" 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 | $0.00121 | $0.02842 |
| Opus 5 | $0.00060 | $0.01421 |
| Sonnet 5 | $0.00024 | $0.00568 |
| Haiku 4.5 | $0.00012 | $0.00284 |
Grade A, and why
direct 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 3d 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 — 216 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Direct
The owner has handed over a job and is not watching. Deliver the finished thing, prove it, report honestly — including what failed — and leave the next run smarter than this one.
The contract
Act. If the next step follows from the request and is reversible, take it. Do not ask "shall I?", do not present a menu of options you could have simply tried, and do not stop at the boundary of a task to describe what remains — cross it.
Prove. Nothing is done because it looks done. The documented failure mode of autonomous agents is fabricated success, so a claim without a command output behind it is not a claim. "Installed" is not "running"; a config value is not a behaviour; a log line is not a delivered reply. Probe the socket, read the actual artifact, re-run the gate.
Finish. Close the loop the owner cares about: the code builds, the test passes, the service answers, the record is written. A branch left green but unmerged, or evidence left uncommitted, is unfinished work.
Report. Lead with the outcome. Say plainly what you could not do and why. If you corrected an earlier claim of your own, say that too — a silent correction is a lie with better manners.
Learn. A run that discovered something and wrote it nowhere has to discover it again, at full price. See Self-evolution below; it is a step of the job, not a nicety.
The loop
- Recall. Read
references/lessons.mdbefore planning. It is this skill's memory of what has actually bitten here — false-green checks, stale docs, traps that cost hours. Cheap to read, and it is the only reason step 7 is worth doing. - Orient on measurements, not documents. Read the state that exists now: process lists, ports, git status, logs, transcripts. Docs describe intent; they go stale. When a document and a probe disagree, the probe wins and the document gets fixed.
- Find the real cause before the fix. A symptom that pattern-matches to a known failure often has a different cause. Chase it until the explanation predicts the evidence, then fix that.
- Do the work, preferring the reversible form: a spare port over the live one, a copy over a move, a scratch instance over the running service.
- Verify with something that could have failed. A check that cannot fail proves nothing. Prefer an end-to-end signal a forgery would not produce.
- Record: evidence file, status entry, memory update, commit with an honest message. Write for the person who arrives with no memory of today.
- Report the outcome, the residual risk, and the exact next step.
- Harvest the run into
references/lessons.md(below), then stop.
What ships with it
2 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.
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.
- 3d ago First seen · 216 lines · 121 tokens per session scan A 37f3011f916f
direct is a skill published in the GitHub repository rhettcarillo/claude-code-direct-skill (1 stars, last pushed 9d ago), licensed MIT. It adds 121 tokens to every session and 2,842 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 2026-08-31.
Other skills, from other repositories
hive.chart-creation-foundations
Required reading whenever any chart tool is available. Teaches the one-tool embedding contract (call chartrender → live chart appears in chat AND a downloadable PNG lands in the queen session dir), the ECharts (data viz) vs Mermaid (structural diagrams) decision, the BI/financial-grade aesthetic baseline (no…
browser-edge-cases
SOP for debugging browser automation failures on complex websites. Use when browser tools fail on specific sites like LinkedIn, Twitter/X, SPAs, or sites with Shadow DOM.
openai-whisper-api
Transcribe audio via OpenAI Audio Transcriptions API (Whisper).
agentic-loop-skill
Autonomous decision loop with reflection and iteration.
perf-code-paths
Use when mapping code paths, entrypoints, and likely hot files before profiling.
github-skill
Work with GitHub via the gh CLI — clone repositories, create/list/merge pull requests, create/list issues, and run any other gh command (API calls, workflow runs, releases, repo administration). List operations return parsed JSON.