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.
git clone --depth 1 https://github.com/TyrusRC/praetorWrote 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/tyrusrc/praetor/exploit)<a href="https://agentmods.dev/commands/tyrusrc/praetor/exploit"><img src="https://agentmods.dev/badge/commands/tyrusrc/praetor/exploit/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/commands/tyrusrc/praetor/exploit"><img src="https://agentmods.dev/badge/commands/tyrusrc/praetor/exploit.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 | $0.00028 | $0.00706 |
| Opus 5.5 | $0.00011 | $0.00282 |
| Sonnet 5.5 | $0.00006 | $0.00141 |
| Haiku 4.5 | $0.00003 | $0.00071 |
Grade A, and why
exploit 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 16d 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 — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are running /exploit — an objective-driven red-team on $ARGUMENTS.
Goal: reach the objective via the cheapest high-impact path — quick wins and
low-hanging fruit first — and deliver a full attack-path map markdown, not a
vulnerability checklist. Follow .claude/rules/hunting.md, the
command-engagement and playbook-red-team-web skills, and stay inside the
safety model (benign PoC proves impact; Rules 5–9 HARD; confirm-before-dangerous
on the raw tools).
Phase 0 — confirm (R28/R32, ask once if unstated)
- Parse target/scope from the arguments; if none, ask.
- Confirm the objective (data exfil / account takeover / admin access / specific flag or asset) and the stealth/noise budget — red team is kill-chain to objective, not full coverage.
Phase 1 — quick-win sweep (low effort, high impact FIRST)
Prioritise the cheapest paths to real impact before deep testing:
- default / weak creds on exposed panels,
check_tech_vulns+map_tech_to_cvesfor a known-CVE quick win, exposed secrets (extract_js_secrets,dump_exposed_git,run_trufflehog), exposed admin/debug endpoints, auth bypass (test_login_bypass), SSRF → cloud IMDS creds, IDOR/BOLA on predictable IDs, subdomain takeover. - Network lane where in scope:
run_network_recon→ sanctioned impacket/netexec,crack_hashes→ credential reuse. Rank targets withrank_attack_targets/risk_rank_endpoints; pursue the highest impact-per-effort first (R29).
Phase 2 — exploit to prove impact (benign PoC)
Confirm each win by REACHING the sink — a read marker / version / benign action
proves you could go further (Rule 8). Log every action to the ATT&CK operator
log (record_redteam_action / it is auto-tagged). For a genuinely dangerous
state-changing action, ASK the operator first (the raw tools enforce this).
Phase 3 — build the attack-path map
plan_attack_paths(domain)— beam-searches confirmed findings to high-value objectives, names the missing capability on near-misses, and emits the severing control per chain.propose_chains/engagement_graphfor the lineage.
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.
- 16d ago Changed · +3 lines dba6aaf5daf1
- 17d ago First seen · 49 lines · 28 tokens per session scan A 377dca25767f
exploit is a command published in the GitHub repository TyrusRC/praetor (7 stars, last pushed 7d ago), licensed Apache-2.0. It adds 28 tokens to every session and 706 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-09-29.
Other commands, from other repositories
launch
Parse all configs in the current engagement, dispatch senior-pentester subagents in parallel, run the validation chain, emit findings ready for /review.
start
Start a new firewall-review engagement. Argus greets the operator, asks all six scoping questions in one batch, scaffolds the engagement folder, and waits for configs. No arguments.
report
Render a consolidated customer-review Excel workbook, optional audit-grade PDF, and chain-of-custody manifest from approved findings. Gated on ≥1 approve in feedback.jsonl.
review
Triage findings in chat. Operator sees each finding with evidence, chooses approve / request-changes / other per finding (or batches).
pending
Curator commands for learning-loop candidates. Usage — /pending list | /pending review | /pending promote | /pending reject | /pending hold.
configure-chiliforms
Build a ChiliForms embed (a form Chili Piper generates from a Concierge router) for a customer without a form of their own, and get a ready-to-paste snippet. Read-only.