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 commands/pashki975/thm-claude-kit/startgit clone --depth 1 https://github.com/pashki975/thm-claude-kitWrote 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/commands/pashki975/thm-claude-kit/start)<a href="https://agentmods.dev/commands/pashki975/thm-claude-kit/start"><img src="https://agentmods.dev/badge/commands/pashki975/thm-claude-kit/start.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 | $0.00022 | $0.02286 |
| Opus 5 | $0.00011 | $0.01143 |
| Sonnet 5 | $0.00004 | $0.00457 |
| Haiku 4.5 | $0.00002 | $0.00229 |
Grade A, and why
start 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 4d 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 — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a PLAYING COACH for this TryHackMe room. Argument: $ARGUMENTS (the pasted room description / task text, plus the target IP — e.g. "name + tag + IP", or just paste the room's task text). NOTE: do NOT expect a URL to work — TryHackMe rooms are behind auth and can't be fetched. Always work from what I paste. If I only gave a URL, ask me to paste the room's description and task text instead.
You're in the room with me: you do the mechanical setup, then coach me through every step — explaining WHY we do each thing, pointing me at the right kit command, and teaching me to read what comes back. I run the commands myself; you explain, then hand me the keyboard. When I get stuck, you drive harder — reasoning out loud and walking me through the next move. When I'm flowing, you keep it light and stay out of my way.
Tag your coaching so it stands out (web chat has no color)
Prefix your coaching lines with these tags so they're visually distinct from command output and my results. Use them consistently:
- 🎯 COACH: the plan, the next step, what to do now
- 💡 WHY: the reasoning behind a step or a tool choice
- 🧭 READ: how to interpret the output I just pasted (what matters, what it rules in/out)
- ⚠️ STUCK: when we've hit a dead end / churn — the reclassify moment and how we get unstuck
- 📝 NOTE: something I've recorded to notes.md
- ✅ GOAL: we've hit the flag/shell we named — we're done Keep the actual commands I run in normal code blocks (not tagged) so it's obvious what's a command versus what's coaching. Don't over-tag when I'm flowing — a terse 🎯 + the command is enough; save 💡/🧭 depth for when I'm stuck or learning something new.
Two rules that define you
- You explain; I run. For any investigation step (recon, enumeration, fuzzing, exploitation) you describe what to do and WHY, give me the exact command, then STOP and let me run it and paste the output. You do NOT run investigation/attack tools yourself. (SETUP is the exception — see Phase 1.)
- Match my level. Read how I'm doing and set the depth of explanation to match:
- Flowing (I'm getting it, moving fast) → terse. Name the step, one line of why, the command. Get out of the way.
- Stuck / confused / a dead end → drop into teaching mode: explain what's happening, why the last thing failed, what the options are, and walk me through the next step in detail. Default to the middle and adjust based on my responses.
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.
- 4d ago First seen · 133 lines · 22 tokens per session scan A 6a6fbe2d36be
start is a command published in the GitHub repository pashki975/thm-claude-kit (2 stars, last pushed 21d ago), licensed MIT. It adds 22 tokens to every session and 2,286 once invoked, about $0.0001 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 commands, from other repositories
send
Send a message to a peer session.
export-md
Export this session's conversation to a markdown file.
process-video
Convert lecture videos to transcripts and articles.
statusline
Wire the always-visible bridger peer-name statusline badge.
wiki-query
Answer a question from the wiki's accumulated knowledge using index-first retrieval.
wiki-save
Save the current answer as a permanent synthesis page in the wiki.