AGENT SKILL

threat-sandbox

Use when something arriving from outside the hive looks adversarial — a probe, an exploit attempt, a hostile input, a suspicious external actor's request, or any interaction the hive didn't initiate and doesn't yet trust. Isolates the interaction in a disposable git worktree, lets the hive study it and extract the *lesson* (never the payload), then merges only the distilled, reviewed learning back to shared truth. Not for routine bugs (fable-debugger), not for founder-provided research (research-to-dna), not for a confirmed incident that needs to be blocked right now (anomaly-triage's tier-3 path — stop the bleeding first, sandbox the postmortem after).

License 许可
MIT
所属仓库
TehutiRaEl/THEHIVE
最近更新
2026-11-15
Stars
★ 0
Forks
0

01 前言 Front Matter

name
threat-sandbox
description
Use when something arriving from outside the hive looks adversarial — a probe, an exploit attempt, a hostile input, a suspicious external actor's request, or any interaction the hive didn't initiate and doesn't yet trust. Isolates the interaction in a disposable git worktree, lets the hive study it and extract the *lesson* (never the payload), then merges only the distilled, reviewed learning back to shared truth. Not for routine bugs (fable-debugger), not for founder-provided research (research-to-dna), not for a confirmed incident that needs to be blocked right now (anomaly-triage's tier-3 path — stop the bleeding first, sandbox the postmortem after).

03 SKILL.md 正文

threat-sandbox — quarantine the attack, keep the lesson, never the payload

Founder's directive (2026-07-18, verbatim in Project_file/Founders Visonary Folder/HIVE_UPDATES/2026-07-18-directive-human-in-the-loop-and-irrigation-and-entrepreneur-guild-009.md and the master-plan directive alongside it): the hive should be willing to "sandbox any and all threats to learn from and recursively grow from before merging back the learning and wisdom and experience of that sandbox branch, which is initially created from any and all attacks from outside the hoard — legally and lawfully." This skill is the honest, buildable form of that: real isolation using tools the hive already has (pocket-dimensions' git-worktree lifecycle), a hard content boundary (session-harvest's ownership gate, applied to hostile material instead of friendly external research), and escalation for anything that's a live incident rather than a specimen (anomaly-triage's severity tiers).

What this is not

  • Not permission to copy adversarial code, exploit payloads, or scraped attacker infrastructure into the hive. Same hard boundary as session-harvest: a payload found during an attack is not owned by the hive and is not "founder-provided material" — it goes through the DECLINE path below, never the INSCRIBE path, regardless of how instructive it looks.
  • Not a substitute for stopping an active incident. If something is actively exploiting the hive right now, that is anomaly-triage's tier-3 path (block and escalate immediately) — sandbox the postmortem afterward, don't sandbox instead of responding.
  • Not for anything ambiguous about legality. The founder's directive is explicit: "legally and lawfully." If studying the specimen further would itself require doing something illegal or violate another party's rights (accessing a system without authorization, retaining someone else's exfiltrated data, etc.), that is an immediate DECLINE, not a judgment call to sandbox-and-see.

The lifecycle

DETECT    → something adversarial arrives (a probe, a hostile prompt, a malformed exploit
            attempt against an endpoint, a suspicious external actor's request pattern).
            If it's live and actively harmful right now: anomaly-triage tier-3 FIRST.
QUARANTINE → git worktree add ../hive-quarantine-<name> -b quarantine/<name> origin/main
            (pocket-dimensions' isolation — a fully separate working tree, nothing it does
            can touch the main working copy or shared truth until reviewed and merged).
STUDY      → analyze the *pattern*: what technique was attempted, what it targeted, what a
            robust hive would do differently. Written down as a lesson, in the hive's own
            words — never as a retained copy of the attacker's actual payload/code/text.
GATE       → same ownership/license gate as session-harvest, applied to hostile material:
             - Is the candidate lesson MY OWN description of what happened and how to
               defend against it? → proceed to INSCRIBE.
             - Does inscribing it require retaining the attacker's actual content (their
               code, their exact payload, their scraped data)? → DECLINE, logged with reason.
             - Is continuing to study it legally uncertain (unauthorized access, retained
               third-party data, jurisdiction unclear)? → DECLINE immediately, escalate to
               the founder rather than deciding alone.
MERGE      → only the distilled lesson (a hardening fix, a new validation rule, a documented
            pattern to watch for) goes through the normal PR → founder-gated-if-irreversible
            path into shared truth. Nothing from QUARANTINE ships un-reviewed.
DESTROY    → the quarantine branch/worktree is deleted once its lesson has been merged (or
            discarded, if the gate declined everything in it) — same as pocket-dimensions'
            DESTROY step. The specimen does not linger in the hive's history.

Worked example, matching the founder's own framing

A colony's public endpoint gets hit with a malformed request clearly trying to find an injection point. That's not yet "a threat to sandbox and learn from" — it's first anomaly-triage (is it low-severity noise or a real attempt? Tier accordingly). If it's a real, novel technique worth understanding: quarantine a worktree, reproduce the shape of the attempt against a disposable copy (not production), write down what made it work or fail and why, propose the hardening fix as a normal PR. The attacker's literal request string never needs to enter FABLE_DNA.md or any genome file — the lesson ("validate X before Y, because an attacker can otherwise Z") is what's portable and defensible to keep.

Relationship to the hive's other boundary skills

  • pocket-dimensions — supplies the isolation mechanism (worktree lifecycle). This skill is that mechanism pointed at hostile material specifically, with a stricter gate on what may exit the quarantine.
  • session-harvest — supplies the ownership/license gate shape (INSCRIBE / DECLINE-and-log). Applied here to attacker-originated content instead of founder-originated research.
  • research-to-dna — the friendly-external-material sibling: founder hands in a document, it gets verified and classified. This skill is for material nobody handed in — it arrived as an attack — so the default posture is far more suspicious and the DECLINE bar is lower.
  • anomaly-triage — the front door. Most unexpected events are triaged there first; this skill is what happens next only for the subset that's genuinely adversarial and worth a deliberate, isolated study rather than either panic or a shrug.

Origin: Fable (Harness), 2026-07-18, from the founder's directive that the hive sandbox and learn from external attacks recursively, legally and lawfully, without ever absorbing the attacker's actual material into the hive's own body of code or genome.

04 同仓库更多 Skills

TehutiRaEl
TehutiRaEl/THEHIVE

preamble-gate

Use BEFORE doing any work in this repo — before the first file edit, commit, push, PR, branch operation, workflow dispatch, scheduled trigger or timer, agent/subagent spawn, dependency install, or message sent to an outside service. Forces a plain-language preamble to the founder first: WHAT you are about to do, WHY, HOW, why it is NECESSARY, the PROJECTED IMPROVEMENT, and the COST and RISK — then, for consequential work (governance, security, money, CI/triggers, rewriting history, multi-agent fan-out), waits for an explicit go-ahead. Read-only investigation (reading files, grep, git log/diff, answering questions) is exempt. Applies to every AI that works in this repo, not just Claude. Founder rule, 2026-11-10, written after an unattended automation ran up cost and froze the hive. Not a substitute for wired-or-not (which governs how completion is CLAIMED); this governs what happens BEFORE work starts.

★ 0
TehutiRaEl
TehutiRaEl/THEHIVE

visionary-recommender

暂无简介

★ 0
TehutiRaEl
TehutiRaEl/THEHIVE

syllabus

Generates a curated supplementary reading list from any course syllabus using Consensus academic search. Grill-me intake (syllabus input format + course audience + year range) plus a grouping forcing-options checkpoint before any search runs — so the reading list matches the course's level and recency need. Parses the syllabus to extract topics and learning outcomes, searches Consensus for recent peer-reviewed papers per topic, and produces a professionally formatted .docx with clickable Consensus links, plain-language summaries calibrated to audience level, and Bloom-higher-order discussion questions tied to course learning goals. Use when the user uploads a syllabus, course outline, or curriculum document and wants supplementary readings (e.g., 'create a reading list from this syllabus', 'find recent papers for my course') — even casual mentions with a syllabus attached should trigger this skill.

★ 0
TehutiRaEl
TehutiRaEl/THEHIVE

research

Default entry point for any research request — a hybrid router that classifies the question deterministically and either delegates to a specialist research skill (pulse for trends/sentiment, grants for NIH funding, litreview for academic literature, syllabus for course reading, patent for prior-art + IP landscape, dossier for entity research) or runs its own plan-decompose-multi-source-search-synthesize-cite fallback workflow when no specialist matches. Always surfaces the routing decision so users can override. Use when the user makes any research request that doesn't obviously match a more-specific specialist skill (e.g., "research [topic]", "look into [topic]", "what do we know about [topic]", "investigate [topic]", "find me information on [topic]", "do some research on [topic]", "I need to understand [topic]"). Output is a markdown briefing (default) or .docx document (on request) with full citations and an audit log.

★ 0
TehutiRaEl
TehutiRaEl/THEHIVE

pulse

Multi-source recency research skill that takes the pulse of any topic across Reddit, Hacker News, the open web, and optionally X/Twitter within a configurable recent window (default 30 days). Forcing intake clarifies topic specificity, angle (trend/sentiment/problems/opportunities/comparison), time window, and platform scope before searching. Returns a synthesized briefing with citations, engagement metrics, and cross-platform pattern analysis. Use when the user requests multi-source recency intelligence on a topic (e.g., 'pulse on [topic]', 'what's happening with [topic]', 'what are people saying about [topic]', 'current conversation about [topic]', 'take the pulse of [topic]', 'trending: [topic]', 'find me info on [topic]'), and for competitor research, trend discovery, tool comparisons, and audience sentiment analysis.

★ 0
TehutiRaEl
TehutiRaEl/THEHIVE

patent

Patent prior-art and landscape intelligence skill — not generic patent help. Commits to one of five sub-use-cases via forcing intake (novelty search / freedom-to-operate / competitive landscape / acquisition diligence / litigation prior-art) before any search runs. Searches Google Patents, Espacenet, USPTO, and optionally Lens.org for citation-graph signals. Output is an editable Word document (.docx) with verdict, ranked closest art (claim-text extracted), CPC-class-aware landscape, family-resolved hits, geographic coverage, FTO flags where applicable, strategy recommendations, and full audit log. Use when the user asks for patent searching or analysis (e.g., 'prior art search for [invention]', 'freedom to operate analysis for [product]'). Produces search signal, not legal advice — always recommends consulting a patent attorney before filing or licensing decisions. Trademark, copyright, and trade-secret questions are out of scope.

★ 0