breach
Skill0 starsChanged 59 days agoMITView on GitHub
General HTML presentation layer for fast single-page designs, test reports, status pages, diagrams, or slide-like pages, plus dedicated structured discussion-digest and archive modes. Trigger on HTML artifact requests; requests to create an investigation or evidence wall (调查墙、线索墙、证据墙、侦探墙) from materials; a cognitive map (认知地图、线索地图) described with photos, clippings, pins, connections, or an investigation-wall look; explicit structured discussion digests ("digest this thread"); `$breach --archive`; requests to render an existing dossier material package; or dossier retrieval handoffs for default archive display. A bare request for a cognitive/concept map, ordinary reading, explanation, or text summaries does not by itself request an investigation wall or HTML. Respect an explicitly requested output medium. Content workflows such as conops remain the content owners. Not for native docs, decks, sheets, full websites, or rendered video files.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- simota/agent-skillstheir own version, first seen Oct 23, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- simota/agent-skills★ 99same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- acquisitionSkill
- after-actionSkill
- cartographic-reconSkill
- chartSkill
- codifySkill
- conopsSkill
- coverSkill
- dossierSkill
- executeSkill
- go-nogoSkill
- herdrSkill
- netease-mailSkill
- neutralizeSkill
- paperworkSkill
- passdownSkill
- repairSkill
- sanitizeSkill
- taskingSkill
- tracebackSkill
- warroomSkill
- x-likes-digestSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 0 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
Test report, HTML → caller's report content + Breach General page rendering.
- Very specific: line 90
Before drafting or materially restructuring technical narrative in the selected mode, invoke paperwork.
- Very specific: line 108
Do not force HTML for a plain request to "explain simply".
- Very specific: line 123
When invoked by another skill (e.g., repair), the caller provides the output path.
- Very specific: line 138
Claude Code: use frontend-design for generation, passing the observed artifact brief, inherited constraints, scoped overrides and relevant …
- Very specific: line 161
Stop before rendering → render=pending, never pass.
- Very specific: line 163
Only set expected-route when user/caller intent unambiguously requires that branch.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Quick Page
Fast, single-page HTML artifacts for daily dev communication.
Output Boundary
Breach is the general HTML presentation layer, not a mutually exclusive topic
category alongside conops. Designs, test reports, and discussions can all use
HTML. The caller owns the task's content contract and facts; paperwork owns
technical writing discipline when authoring is needed.
The dedicated Discussion digest mode additionally organizes an explicitly requested structured digest using its bundled schema, then renders deterministic HTML. This request can default to HTML without a separate format instruction; ordinary discussion reading, explanation, or text summaries do not enter this mode or automatically produce HTML.
- Technical design, Markdown →
conops. - Technical design, HTML →
conopscontent + Breach General page rendering. - Test report, HTML → caller's report content + Breach General page rendering;
HTML alone does not invoke
conops. - Native document, presentation, or spreadsheet requested → use the requested OpenAI Template or native artifact capability.
- HTML page, HTML slide-like page, or discussion digest requested → use breach.
- Another workflow may define content and evidence first, then call breach only for HTML presentation. Approved prose need not be a persisted Markdown file.
- For
conopsoutputhtmlorboth, use General page mode and preserve its 14-section content contract.htmlrequires no Markdown artifact;bothderives HTML from the Markdown maintenance source. Conops validates the final HTML with its own quality gate; Breach rendering and visual checks still apply.
A requested HTML storyboard, video review page, or player belongs to General page. A requested rendered film belongs to an available video-production capability; pass the project DESIGN.md, selected direction, content, and delivery requirements. If none is available, identify the gap without substituting an HTML animation for a video. Do not invoke a hypothetical skill.
Core Rule
Deliver efficiently while preserving the chosen art direction. A generic page that erases the requested style is incomplete even if it renders correctly.
Choose one mode:
- General page — resolve the artifact direction and inherited constraints, then use website and layout references only where needed under the shared collaboration rules; load the interaction reference for controls or requested motion.
- Discussion digest — use the bundled schema and renderer; resolve an artistic direction and use an artifact-specific template when appropriate.
- Archive (
--archive) — render only a registered, dossier-gate-validated standard package with the bundled archive renderer/template and its reference-backedassets/archive-DESIGN.md. Material acquisition and conclusions remain upstream; no automatic dossier creation.
For a General page or Discussion digest with an explicitly requested new artistic style, run the existing Cover style router before rendering, inspect its status and record the actual receipt in provenance, as specified in General page. Do not infer that reading the reference or naming a style is equivalent to running it. If the router is unavailable, follow the documented fallback and disclose it. A previously routed unchanged direction can reuse its existing receipt.
Minimal entry
Choose one artifact mode and read only its route:
$breach --archive [manifest.json]or render an existing dossier material package → Archive. Require the dossier manifest and its identity catalog; the renderer runs dossier gate before rendering. An omitted input must resolve to one explicit registered package in context; otherwise ask for it. Reject raw documents or ad-hoc manifests; never standardize them inside breach. “调取XX的档案” first belongs to dossier for read-only lookup; dossier then calls Archive by default with the registered manifest, catalog and a separate output path. Breach does not perform the lookup or modify the source dossier.- Investigation/evidence wall from supplied materials, or a cognitive map explicitly described in that visual style → General page; default to interactive HTML when no medium is specified. Keep fact, allegation, inference and adaptation distinct; label relationship meanings and preserve source links and gaps. Apply the image-led exploration guidance when useful; image generation is not a substitute for verified evidence. A standalone request to retrieve an existing dossier still follows dossier → Archive unless the user asks for the wall presentation.
- HTML status page, diagram, slide-like page, PR/incident page → General page for reference collaboration, scoped tokens, optional chart gate, and provenance.
- Multi-party email, issue, PR, chat, or forum discussion digest → Discussion digest for schema and deterministic renderer.
For each mode, apply the writing discipline below before authoring new technical narrative. A supplied approved text rendered faithfully needs no re-authoring.
Content Routing
Apply the content profile after choosing the artifact mode; it does not create another renderer:
Writing discipline
Before drafting or materially restructuring technical narrative in the selected mode,
invoke paperwork. It owns the content model, reader task,
information relationships, evidence calibration, and prose quality; Breach owns
HTML structure, visual tokens, rendering, and provenance. Follow that skill's
required calibration-corpus step before writing.
Apply this discipline when generating page copy, explanations, conclusions, decision rationale, status narratives, or discussion-digest synthesis. If the caller supplies approved prose and asks only for faithful HTML rendering, keep the prose unchanged and skip re-authoring. Deterministic rendering never grants Breach permission to invent or strengthen facts.
- Default — preserve the caller's normal level of detail and structure.
- ELI5 — select when the caller marks
content_profile: eli5or the user makes an explicit beginner-level request. Keep the General page renderer, but use one idea per section, short sentences, and concrete analogies labelled as analogies. Preserve evidence, qualifiers, uncertainty, and provenance; never invent a fact to make an explanation simpler. - Do not force HTML for a plain request to "explain simply". If no HTML artifact
is requested, answer inline; use
visualizeonly when a visual materially improves understanding.
Output Location
HTML artifacts write to the current repo working directory, NOT to the source material directory. Default output paths by context:
- Defect fix plans →
.proposal/repair/<bug-id>/index.html - Design proposals →
.proposal/<topic>/index.html - Research findings →
.research/<topic>/index.html - Archive →
.proposal/archive/<dossier-id>/index.html, outside the read-only source package. - General artifacts → caller-specified path, defaulting to
.proposal/if unspecified
When invoked by another skill (e.g., repair), the caller provides the output
path; breach accepts it and writes there. Never write HTML into
/media/yhr/2T/yunxiao/ or other Phase 0 scraped data directories.
Canon 输出边界
读取共享契约:/home/yhr/.agents/repos/agent-skills/references/canon-output-contract.md。
- HTML 页面是 artifact,仍写在调用者指定的
.proposal/、.research/、queries/或其他 repo-local 路径。 - Discussion digest 的 JSON 是可审计中间产物;决议、争议和 action items 可提升到 Canon
decisions/、tasks/、patterns/或 update-card。 - breach 不主动复制 HTML 到 Canon;Canon 默认只记录 absolute path、HTTP URL、页面类型和它支持的 task/decision/incident。
- 如果页面承载长期结论,创建或更新
/media/yhr/2T/Canon/raw/update-cards/<date>-breach-<topic>.md,或让调用方 skill 负责 promotion。
Agent-Specific
- Claude Code: use
frontend-designfor generation, passing the observed artifact brief, inherited constraints, scoped overrides and relevant layout patterns. - Codex / Pi: generate HTML directly from the same resolved artifact brief and scoped tokens; references inform rather than replace its direction.
Routing telemetry
At entry, retain the trigger category and entry form. After the final local
verdict, emit one event following
the shared routing contract.
Use route=general-page|discussion-digest|archive|native-handoff|inline-answer|unresolved.
A route is the branch actually entered, even if it later fails. Record
content-profile=default|eli5, prose=author|faithful, and the selected style
source category. For General page, also record chart/Hairline decisions after
resolving availability and integration; for digest, record the actual style
source and use charts=omit, hairline=omit. Archive uses style=archive-template, prose=faithful, and the same omitted chart/illustration decisions; record actual render/visual checks. It preserves existing relationship views, not new analytical content.
Record actual step states for paperwork, layout, style, render, charts,
hairline, visual-check. pass requires the relevant completed work or check;
selection alone is pending. Faithful prose → paperwork=skipped; digest →
layout and style reflect the actual selected template and style work.
Modules omitted by decision → skipped; unavailable requested modules with a
working native fallback → module skipped, rendered fallback verified separately.
Stop before rendering → render=pending, never pass.
Only set expected-route when user/caller intent unambiguously requires that
branch. Never infer expectation from your own selected route. Attach generated
HTML/JSON and available check references via artifact/gate. Native/inline
handoff emits skipped for Breach artifact production, with the actual route.
Telemetry failure remains a warning; nested skills emit their own events.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

