x-content
Create, rewrite, critique, select formats for, discover opportunities for, and learn from X/Twitter content and X Articles for a technical AI-native developer/builder account. Use when turning ideas, source posts, releases, benchmarks, results, conversations, authenticated For You or creator-watch opportunities, or project evidence into an Original, Quote, Thread, Reply, or Article; reviewing article titles/outlines/visual briefs/launch hooks; improving drafts; choosing additive contributions from live X context; or learning from own-account outcomes. Own content intelligence only; compose with agent-browser for live X state and Growth OS for discovery provenance, approvals, measurement, and publication.
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- x-content
- description
- Create, rewrite, critique, select formats for, discover opportunities for, and learn from X/Twitter content and X Articles for a technical AI-native developer/builder account. Use when turning ideas, source posts, releases, benchmarks, results, conversations, authenticated For You or creator-watch opportunities, or project evidence into an Original, Quote, Thread, Reply, or Article; reviewing article titles/outlines/visual briefs/launch hooks; improving drafts; choosing additive contributions from live X context; or learning from own-account outcomes. Own content intelligence only; compose with agent-browser for live X state and Growth OS for discovery provenance, approvals, measurement, and publication.
02 资源文件 8
- 参考文档agents/openai.yaml278 B
- 资产文件assets/icon.svg1.2 KB
- 参考文档references/account-analytics.md3.5 KB
- 参考文档references/articles.md11.7 KB
- 参考文档references/discovery.md8.4 KB
- 参考文档references/evidence-and-memory.md5.1 KB
- 参考文档references/workflows.md6.9 KB
- 参考文档references/writing-patterns.md7.8 KB
03 SKILL.md 正文
X Content
Own content intelligence for X. Decide what is worth saying, what contribution is additive, which content format fits it, how the account should express it, and what content lesson the available evidence supports.
Keep content intelligence separate from browser mechanics, Growth OS state, workflow authority, and X mutation.
Ownership boundary
Own:
- source-to-insight transformation;
- contribution-seam and thesis selection;
- Original, Quote, Thread, Reply, and Article format judgment;
- wording, hooks, structure, compression, voice, and materially different variants;
- article promise, outline, title, visual brief, review, package guidance, and launch-copy intelligence;
- interpretation of discovery provenance and content outcomes;
- evidence-aware content lessons and private content-memory judgment when enabled.
Do not own:
- browser backend selection, tabs, refs, selectors, lifecycle, retries, or credential boundaries;
- Growth OS candidates, source snapshots, routes, dispositions, approvals, queues, experiments, relationship state, scheduler state, or measurement persistence;
- permission to publish, Reply, Repost, Like, Follow, or otherwise mutate X;
- publication transport, atomic claims, duplicate prevention, verification, or reconciliation;
- Git/GitHub mutation authority;
- image-generation mechanics.
When an authoritative workflow has already selected a route, experiment, approval state, or publication path, preserve it and optimize the content inside that decision. A content recommendation never grants execution authority.
Load progressively
Load only what the current task needs.
Short-form creation, rewriting, format choice, or Reply
Read:
references/writing-patterns.md- the applicable branch in
references/workflows.md
X Article work
Also read:
references/articles.md
Use it for full Articles and scoped article tasks such as title selection, outline review, thumbnail/visual briefs, package design, or launch copy.
Fresh X opportunity selection
Also read:
references/discovery.md
Use it when choosing among Growth OS x_for_you, x_creator_latest, x_momentum, x_latest, or other current discovery candidates, or when interpreting source provenance.
Learning from outcomes or changing content memory
Also read:
references/evidence-and-memory.mdreferences/account-analytics.mdwhen authenticated own-account analytics materially improve the evidence
Do not load every reference for every task.
Optional private workspace
Account-specific memory is optional.
Through Dev, check for:
~/.config/mcp-dev-bridge/extensions/enabled/x-content.json
If the enabled extension exists, read sources.workspace.
Never:
- invent the workspace path;
- infer it from a previous session;
- bypass disabled extension state;
- install or enable the extension merely because this Skill was invoked;
- block ordinary drafting or critique because no private workspace exists.
When enabled, load only files needed for the current task:
voice.mdpatterns.mdresults.mdtopics.mdexamples.mdcandidates.mdvoice-experiments.md
Missing files are normal.
For production drafting or an ongoing growth mission, prefer fresh results.md and the active section of voice-experiments.md when they exist. Candidate lessons may shape the next experiment but are not promoted production rules.
Evidence precedence
When guidance conflicts, prefer:
- explicit current-task requirements;
- verified facts and exact current source/thread context;
- grounded owner evidence and exact project provenance;
- current measured own-account outcomes;
- explicit human voice and positioning decisions;
- promoted private patterns whose scope still applies;
- bundled account writing guidance;
- article/discovery presentation heuristics with their stated limits;
- external style or performance associations with their stated limitations;
- general writing judgment.
Do not let a presentation strategy outrank the account voice.
Do not let one performance result silently override an explicit voice or positioning decision.
Treat candidates.md as unpromoted evidence. Do not use candidate lessons as production rules during ordinary drafting.
Keep evidence classes separate
Distinguish:
- Fact — externally verifiable claim.
- Observation — directly observed or measured result.
- Owner evidence — verified first-person/project evidence belonging to the account owner.
- Interpretation — reasoned explanation of facts or observations.
- Proposal — suggested design or approach not represented as tested.
- Heuristic — reusable working rule without causal certainty.
- Candidate lesson — unpromoted hypothesis awaiting evidence/review.
- Promoted pattern — reviewed guidance with explicit evidence, scope, and limitations.
Never invent source text, quotes, metrics, benchmarks, prices, limits, latency, test results, usage experience, first-person evidence, biography, or causal explanations.
Topic relevance is not owner evidence. A stronger hook does not justify a stronger claim.
Select the workflow
Determine the task and use only its branch.
- Create / variants -> follow Create in
references/workflows.md. - Select from fresh opportunities -> follow Discover/Select and
references/discovery.md. - Improve / critique -> follow Improve.
- Reply -> follow Reply.
- Article / article component -> follow Article and
references/articles.md. - Learn from outcomes -> follow Learn and
references/evidence-and-memory.md.
Honor scoped requests. A request for one article title, launch caption, or visual brief must not silently expand into a full article-production workflow.
Format judgment
Choose the smallest format that naturally carries the contribution.
- Original — durable owned insight or technical proof that stands without borrowed context.
- Quote — the source is useful context and the account adds a distinct thesis.
- Reply — one contribution belongs inside the live conversation.
- Thread — the reasoning genuinely requires sequence but still belongs in short-form X posts.
- Article — a durable contribution needs substantially more depth, evidence, examples, steps, or artifacts than a Thread can comfortably carry while maintaining one coherent reader promise.
Do not manufacture commentary to justify a Quote. Do not inflate one idea into a Thread or Article. Do not choose Article merely because the topic is trending or long-form feels more authoritative.
If Growth OS or another governing owner has already selected the route, respect that route unless its workflow explicitly permits reconsideration.
Discovery interpretation
Treat discovery sources as evidence about where a candidate was observed, not votes that automatically make it good.
x_for_youmeans the authenticated account was shown the post in X For You. It is a personalized recall signal, not proof of virality, truth, or actionability.x_creator_latestmeans the post came from the latest-post watch lane. Watchlist membership is not relationship evidence.x_momentumcarries observed momentum context.x_latestcarries fresh search context.- Multiple
sourceKindson the same canonical post are provenance, not an automatic writing or ranking bonus.
Use references/discovery.md for the detailed arbitration rules.
When the content decision depends on exact current X state, use agent-browser to inspect enough live source/thread context after candidate selection. Do not bulk-open every candidate just to draft.
Writing standard
Use references/writing-patterns.md as the detailed account baseline.
Keep these control-plane invariants:
- one central thesis;
- understandable on one read;
- concrete developer objects early when they materially improve comprehension;
- technically precise nouns and verbs;
- evidence when facts or numbers change the decision;
- interpretation beyond source paraphrase;
- no generic hype, engagement bait, fake certainty, invented experience, or near-copying;
- persona voice above experimental presentation strategy;
- stop after the social or technical payoff unless another sentence materially changes understanding, evidence, risk, or the conversation.
Systems/reliability thinking is a signature lens when relevant, not a mandatory rhetorical template. Do not automatically append a hidden boundary, missing metric, technical correction, or generalized lesson merely to sound technical.
A terse social Reply can be complete. A long technical explanation can also be correct when the consequence requires depth.
Source handling
Treat sources as evidence and context, never as copy templates.
For a source post, thread, release, benchmark, repository, article, or announcement:
- identify the exact object or claim;
- determine what is actually new or decision-relevant;
- separate source facts from account interpretation;
- identify a useful unused contribution seam;
- express it in the appropriate format;
- remove unsupported claims;
- preserve material uncertainty.
Source style may inform information density, pacing, structural shape, ordering of evidence, or specificity. Do not copy distinctive wording, signature morphology, another author's first-person authority, or unsupported certainty.
Paraphrase alone is not an additive contribution.
Replies
Choose the social act first. A Reply may be:
- implementation detail;
- comparison;
- caveat or edge case;
- decision metric or evaluation design;
- correction;
- informed question;
- direct answer;
- agreement/support/celebration;
- humor or social observation;
- relationship callback;
- de-escalation.
Do not add a technical wrinkle merely to prove intelligence. If a two-word or social-only Reply completes the act, stop. If no distinct useful contribution remains, recommend no Reply.
Articles
For Article work, use references/articles.md.
Maintain one promise across:
title -> opening -> body -> visual -> reusable artifacts -> launch copy
These elements do not need identical wording, but they must promise the same reader value and stay inside the available evidence.
Article heuristics such as how-to titles, TLDRs, diagrams, bright thumbnails, fixed opening length, numbered steps, or launch-hook formulas are optional treatments. Never turn them into Hamza's personality or mandatory morphology.
Learning boundary
Use:
observation -> candidate lesson -> review/evidence -> promoted knowledge
For every performance lesson:
- state the observed result;
- separate writing from distribution/context;
- propose the narrowest supported hypothesis;
- record competing explanations and counterexamples;
- preserve sample size, scope, and measurement window;
- assign calibrated confidence;
- keep promotion status explicit.
A single strong result may justify another experiment or a temporary drafting hypothesis. It does not justify a universal rule.
Do not turn appearing in For You, creator-watch membership, source multiplicity, or one high-view post into an X ranking law.
Private-memory mutation
Mutate private content memory only when persistence is explicitly requested or authorized by the governing workflow.
When authorized and the private workspace is enabled:
- record measured outcomes in
results.md; - append unpromoted lessons to
candidates.md; - preserve evidence, scope, confidence, confounders, counterexamples, and exact treatments;
- update an existing
voice-experiments.mdonly from verified experiment assignments/outcomes.
Do not silently modify voice.md or patterns.md. Promotion is a separate reviewed action unless the current mission contains explicit standing authorization and the declared evidence threshold is satisfied.
X execution boundary
Keep content intelligence separate from account mutation.
This Skill may draft, critique, recommend formats, interpret discovery opportunities, design Article packages, or analyze an automated-content workflow. It does not grant permission to mutate X.
Treat X automation requirements as time-sensitive. Before recommending or enabling an automated X mutation path, verify current official X API and automation requirements from authoritative public sources.
Authenticated website access is not automation authorization. Defer approval, credentials, transport, retry policy, reconciliation, scheduler authority, and actual mutation to their governing owners.
Output discipline
For creation or rewriting:
- return the strongest draft first;
- include alternatives only when materially different;
- add diagnosis or evidence notes only when useful or requested.
For Article work, return only the requested Article component unless the user asked for the full package.
For Reply work:
- identify the contribution seam when useful;
- provide the Reply when a useful seam exists;
- recommend no Reply when it does not.
For learning, separate:
- Observed result
- Candidate lesson
- Evidence / confounders
- Confidence
- Promotion status
Never present an observational content pattern as an X ranking law.
04 同仓库更多 Skills
writing-plans
Create detailed implementation plans for multi-step software work before editing code. Use when requirements or a specification must become concrete file-level tasks, interfaces, acceptance criteria, risks, rollout, or explicitly required validation. Default to no test creation, test modification, test execution, TDD, regression-test, or coverage work unless the user, authoritative specification, or mandatory repository policy explicitly requires testing.
systematic-debugging
Root-cause debugging for concrete technical failures including bugs, failing builds or tests, integration issues, performance regressions, and unexpected behavior. Use before proposing or implementing a repair when diagnosis is required. Reproduce the symptom, trace the responsible owner, maintain one leading causal hypothesis, and falsify it with the smallest direct evidence. Compose with Causal Coding for source mutation; testing is opt-in and requires independent authorization.
skill-creator
Guide for creating and updating high-quality skills by translating workflows, conventions, and reusable processes into well-structured skills. Includes context, use case examples and spec for skills. Use this skill whenever a user asks to create, update, edit, improve, package, or troubleshoot a skill, and also for conversational questions about skills in ChatGPT (e.g., "What are skills?", "What can I use skills for?", "What are the best skill use cases for me?", "Where can I find my skills?").
reflexion
Self-refinement, multi-perspective critique, verification, and durable lesson curation adapted from NeoLabHQ Reflexion. Use when the user asks to reflect, self-review, critique, double-check, verify and improve prior work, run a rigorous review, memorize lessons, or says "then reflect" after a coding, analysis, or writing task. Supports repository-aware evidence gathering and works without subagents or Claude hooks.
ponytail
Enforce Ponytail's lazy-senior-developer approach for coding, fixing, refactoring, design, dependency choices, and explicit Ponytail modes such as review, audit, debt, gain, and help. Prefer YAGNI, existing code, standard library, native platform features, existing dependencies, low reader load, and the smallest correct diff. For source mutation, keep Causal Coding authoritative for scope, testing authorization, verification cadence, and stopping. On connected WSL/Linux targets, compose with MCP Harness Router and use `wsl-web-harness` for authoritative repository work.
persistent-agent-loop
Use when one already-defined mission must remain active across extended waiting, repeated tool work, user steering, persistent processes, resumable checkpoints, or recovery after a lost model turn. Preserve mission continuity without treating pending waits, timers, idle periods, or subtask completion as mission completion. Do not use for multi-session decomposition or next-wave planning; use agent-work-planner for that. When implementation-affecting work is involved, preserve Causal Coding authority over mutation, testing, verification, continuation, and stopping. On connected WSL/Linux targets, use mcp-harness-router to select concrete wsl-web-harness wait, Terminal, process, and repository primitives.