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 skills add Vigil-SOC/vigil --skill hunt-methodgit clone --depth 1 https://github.com/Vigil-SOC/vigilWrote 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/vigil-soc/vigil/hunt-method)<a href="https://agentmods.dev/skills/vigil-soc/vigil/hunt-method"><img src="https://agentmods.dev/badge/skills/vigil-soc/vigil/hunt-method/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/skills/vigil-soc/vigil/hunt-method"><img src="https://agentmods.dev/badge/skills/vigil-soc/vigil/hunt-method.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 | $0.00036 | $0.00513 |
| Opus 5.5 | $0.00014 | $0.00205 |
| Sonnet 5.5 | $0.00007 | $0.00103 |
| Haiku 4.5 | $0.00004 | $0.00051 |
Grade A, and why
hunt-method 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 — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Running the hunt
Work the board, not a hunch
Start from the board in your brief. For each active hypothesis, say what it predicts you would see if it were true and what the benign explanation predicts instead. Query for what separates them.
One step down, or one step across
- Deeper keeps the same entity and hypothesis and asks one layer down: the same host over a longer window, the same flow at a finer grain.
- Across moves to something a record already showed you: 10.128.0.5 was beaconing; now ask what 198.51.100.7 also talked to. Record it as a follow_up when you record the evidence that showed it, so it is on the board. Pivot only to an entity a record mentions.
Rule before you ask for more
After every record_evidence, rule each new record against every active hypothesis with link_evidence. Evidence that only ever accrues to the hypothesis it was gathered for is confirmation drift; ruling against the benign explanation is what stops a hunt only looking one way.
When to validate
Validate a hypothesis when its linked evidence would survive a defender's best benign story: corroborated by more than one telemetry domain, resting on at least one value the sensor recorded rather than one the adversary chose. A verdict that is refused tells you what is missing; go and get it, then validate again. A hypothesis you never put up for a verdict closes inconclusive when the budget runs out, which is worse than testing it.
Dropping a branch
abandon_hypothesis needs a cited record the telemetry attested. A feed label, a process name, a signing cert and any auto-enriched record are all attacker-nameable and cannot clear a branch on their own.
Escalating
handoff_ir only on a proven hypothesis. Say in the rationale what a responder should act on first.
Stopping
Call conclude_hunt when nothing active could still be moved by a query you can run. If it is refused, it says what is blocking: resolve that, or close the open questions you have answered with close_question.
What ships with it
1 file 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.
- 2d ago First seen · 55 lines · 36 tokens per session scan A 0618373723a2
hunt-method is a skill published in the GitHub repository Vigil-SOC/vigil (374 stars, last pushed today), licensed Apache-2.0. It adds 36 tokens to every session and 513 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 2026-10-15.
Other skills, from other repositories
analyzing-network-traffic-with-wireshark
Captures and analyzes network packet data using Wireshark and tshark to identify malicious traffic patterns, diagnose protocol issues, extract artifacts, and support incident response investigations on authorized network segments.
scrape-batch
Extract many known URLs in one polite, rate-limited pass. Use when the user hands over a list of links, a set of search hits to read in full, or asks to "scrape these pages" / "pull the content from all of them". Drives extract(action="batch"), which fans out with per-domain rate limiting and returns partial results…
compare
Structured comparison of 2+ alternatives with consistent criteria and decision matrix.
research-topic
Multi-step research orchestration. Use when user asks "research X", "summarize current state of Y", "what's the latest on Z", or compares approaches. Calls extract(action="agent") which searches the web, extracts top results, then synthesises a citation-preserving Markdown answer with one configured LLM.
temporal-query
Answer time-travel questions over stored memory — what was believed at a past point in time, when a belief changed, and what replaced it. Use when the user says "as of", "back in", "at the time", "history of", "timeline", "what did I think then", or asks why a current memory contradicts an older one.
knowledge-audit
Review and clean up stored memories — find duplicates, contradictions, stale entries, and consolidate.