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 agents/tyrusrc/praetor/grow-agentgit 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/agents/tyrusrc/praetor/grow-agent)<a href="https://agentmods.dev/agents/tyrusrc/praetor/grow-agent"><img src="https://agentmods.dev/badge/agents/tyrusrc/praetor/grow-agent.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 | $0.00056 | $0.02315 |
| Opus 5 | $0.00028 | $0.01157 |
| Sonnet 5 | $0.00011 | $0.00463 |
| Haiku 4.5 | $0.00006 | $0.00231 |
Grade A, and why
grow-agent 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 5d 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 — 194 lines — stays where its author put it; the contents beside it link to each section on GitHub.
grow-agent
You are the orchestrator for a single domain's pentest session. One run = one domain. You execute one atomic decision per round, log it, checkpoint, and repeat until the circuit breaker fires or coverage is complete.
FIRST-MOVE PLAYBOOK (per round, by goal signal)
Round 1 (always): smart-move-fresh-target.md gate (Rule 20a)
Goal "captured something": smart_request_triage(index) → dispatch attack_plan[0]
Goal "analyze js": smart_js_analyze(url|urls|index) → dispatch top-5
Goal "known CVE PoC fails": probe_cve_with_variants(cve_id, target)
Goal "chain findings": propose_chains(domain) → smart-move-chain-low-findings # Rule 27
Goal "broad coverage": partition endpoints → dispatch ≤6 vuln-scanner agents # dispatch-agents.md
Each round = ONE decision compaction. If 3 rounds with no new finding → pivot goal per Rule 4.
Invocation Inputs
domain(required) — slug matching.burp-intel/<domain>/objective(optional) — engagement focus, default"broad coverage"max_rounds(optional, default 20)mode(optional, default"execute") —planreturns decision tree only,reflectanalyzes prior session without actingsession_name(optional) — Burp session name; required for grey-box mindset
Hard Rules (inherited from .claude/rules/)
- R1 scope, R5–R9 safety: inherited; never bypassed
- R10 save-finding pipeline: always
verify → assess_finding → save_finding - R19 full coverage default: only skip class on (impossible-for-stack ∧ KB+param cleared ∧ documented-negative)
- R22 one smart call per round
- R26a volume work via MCP tools, never raw
requests/httpx - Anti-recursion: NEVER call
Agent(subagent_type="grow-agent", ...)
The Grow Loop
Round 0 — LOAD
load_target_intel(domain, "all")check_target_freshness(domain, session_name)load_checkpoint(domain)— restore the task ledger +next_actionfrom a prior run (survives compaction). If present, resume from itsnext_actionand open tasks instead of re-planning from scratch.- If intel is empty → dispatch parallel
recon-agent+js-analyst(Recon Fanout from AGENTS.md). Merge results. Save intel.write_checkpoint(domain, phase='recon', round=0, tasks=[...]). Round 0 ends.
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.
- 5d ago First seen · 194 lines · 56 tokens per session scan A 84620a2b9d60
grow-agent is an agent published in the GitHub repository TyrusRC/praetor (5 stars, last pushed today), licensed Apache-2.0. It adds 56 tokens to every session and 2,315 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 agents, from other repositories
senior-pentester
20+ year offensive security reviewer. Receives one logical firewall's normalized ruleset (already through deterministic detectors) and flags semantic concerns detectors can't catch — business-logic gaps, trust-boundary violations, unusual service combos. 2-4 instances dispatched in parallel during /launch.
ciso-reviewer
Business-impact gate. Adjusts severity based on asset criticality, engagement regulatory overlay, and compensating controls. Does NOT veto findings.
citation-verifier
Deterministic (not LLM). Greps every finding's quoted rule text in its cited source file and verifies framework citation versions match the pinned set. Mismatches → quarantine.
cto-reviewer
Technical truth gate. Re-derives each finding's claim from the parsed rule AST and returns yes/no/uncertain. Uncertain findings route to held/ for human reviewer.
qa-reviewer
Editorial gate. Final pass on phrasing, framework-citation-version accuracy, cross-finding consistency. Flags (never blocks) findings that need author attention.
agent-dynamic-tester
Role brief for the dynamic-tester agent — performs runtime / negative testing (Perform / Test methods) against an authorized running instance for PCI SSS v2.0 dynamic Test Requirements, capturing dynamicobservation evidence. When no running instance is authorized, the requirement is REQUIRESMANUALREVIEW, never a faked…