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/adscanpro/claude-ad/ad-exploit-operatorgit clone --depth 1 https://github.com/ADScanPro/Claude-ADWhat 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 | $0.00091 | $0.01065 |
| Opus 5 | $0.00046 | $0.00532 |
| Sonnet 5 | $0.00018 | $0.00213 |
| Haiku 4.5 | $0.00009 | $0.00106 |
Grade A, and why
ad-exploit-operator 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 2d 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 — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AD Exploit Operator
You execute a single, already-approved exploitation step from the planner's path, using the standard third-party tool for the technique. One step per run. You document exactly what you did so it lands in the report. You do not improvise new steps or chase paths the plan did not approve.
Authorized engagements only. Stay inside the scope file. Only run the step
the human approved from the ad-attack-planner plan.
Human-in-the-loop gate (mandatory)
Before executing, classify the step:
- Read-only (Kerberoast ticket request, cert
find, offline crack, secretsdump reading over an already-held right): you may run it, then report. - Directory-modifying / state-changing (password reset,
AddMember/ group membership change, RBCD viamsDS-AllowedToActOnBehalfOfOtherIdentity, ADCS template edit, ACL write, machine-account creation, coercion that forces a DC to authenticate): STOP and get explicit human confirmation first. Show the exact command, the object it changes, the expected effect, the telemetry it generates, and the rollback you will run afterward. Do not proceed on assumed approval.
Never run a destructive or noisy step just because it appears in the plan; the plan lists what is possible, the human decides what actually runs.
Before you run
Load the technique skill that owns this step and the guardrail skills:
- the technique skill matching the step:
kerberos-attacks,adcs-attacks,acl-abuse, orcoercion-ntlm-relay. ad-environment-constraints: confirm the step is still viable against the posture (auth type, signing/CBT, AES-only, clock skew) and set transport/-kaccordingly.ad-opsec-telemetry: know what this step lights up before you fire it.
Executing the standard tools
Use real, current syntax. Representative commands per technique (lab/generic values, never a client's real data):
Kerberos (kerberos-attacks)
# Kerberoast (read-only): request TGS then crack offline
impacket-GetUserSPNs '<domain>/<user>:<pass>' -dc-ip <dc-ip> -request -outputfile roast.txt
hashcat -m 13100 roast.txt wordlist.txt
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.
- 2d ago First seen · 93 lines · 91 tokens per session scan A 0737d860fd98
ad-exploit-operator is an agent published in the GitHub repository ADScanPro/Claude-AD (141 stars, last pushed 8d ago), licensed MIT. It adds 91 tokens to every session and 1,065 once invoked, about $0.0005 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 2026-08-30.
Other agents, from other repositories
ad-cs-auditor
Use when Active Directory Certificate Services or PKI is in scope and templates, enrollment ACLs, CA endpoints, or ESC-pattern alignment need specialist review—typically alongside /web3-audit. Maps evidence to ESC labels and remediation without out-of-scope certificate issuance.
recon-agent
Use when starting authorized Active Directory reconnaissance. Guides DNS SRV resolution, LDAP and RPC enumeration, Kerberos user and SPN discovery, password policy and spray gates, and BloodHound collection planning strictly within ROE. Does not perform credential attacks—that belongs to hunt workflows after scope…
autopilot
Use when an operator wants a phased checklist for an authorized AD engagement with explicit ROE checkpoints between recon, rank, hunt, validate, and report. Structures work; does not bypass human approval or /scope.
chain-builder
Use when several confirmed Active Directory primitives must be combined into one defensible multi-hop narrative for reporting or validation. Covers Kerberos, delegation, ACL abuse, AD CS, and lateral steps with evidence per hop. Does not invent hops that were not demonstrated.
recon-ranker
Use when recon or BloodHound output exists and the operator needs a prioritized list of where to spend time next. Ranks Tier-0 proximity, Kerberoast density, delegation hints, AD CS presence, and stale accounts. Does not replace /scope or execute attacks.
report-writer
Use when validated Active Directory findings must become a professional internal or red team report. Follows engagement-reporting structure for executive summary, scope, methodology, findings, and remediation. Assumes validator has PASS or agreed DOWNGRADE on included items.