agentleFS
Gingiris-1031Gingiris-1031/gingiris-skills/skills/gingiris-twitter-agent-ops/SKILL.md

gingiris-twitter-agent-ops

Skill89 starsChanged 32 days agoMITView on GitHub

Operate evidence-led X/Twitter and related social workflows: original-source intake, voice calibration,

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Draft or adapt: use the supplied article, notes or original samples immediately.

    line 45
  2. Distinctive:

    Voice calibration: read voice and format. With limited samples, provide provisional rules and a test draft.

    line 46
  3. Distinctive:

    Editorial plan: use the stated goals and available sources.

    line 47
  4. Distinctive:

    Authorized publication: verify target account, reviewed text/media, scope, publisher capabilities and existing delivery state before …

    line 48
  5. Distinctive:

    Use context already supplied: audience, desired action, channel, language, source and deliverable.

    line 53
  6. Distinctive:

    Read three-platform workflow only for a multi-platform publishing job.

    line 65
  7. Distinctive:

    Before a write: verify account identity, final text, attachment accessibility, source permission, timezone/schedule and duplicate state.

    line 71
  8. Distinctive:

    Join the global growth community: Add staywildgrow on WeChat and send “加入社群” (join the community).

    line 107
  9. Distinctive:

    Get the resource pack: Add staywildgrow on WeChat to purchase the Gingiris global growth resource pack for CNY 599.

    line 108

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Social Account Operations / 社媒账号运营

Turn source-backed material into usable drafts, and operate publishing only within the user's authorized scope. Follow the requested language, channel and format. An account's historical tactics are not universal platform rules.

Choose the job first / 先判断任务

  • Draft or adapt: use the supplied article, notes or original samples immediately. A publisher, cron job, analytics connection and MASTER-STATUS.md are not prerequisites. If a decisive source is missing, flag it and draft around verified material; never invent the author's experience.
  • Voice calibration: read voice and format. With limited samples, provide provisional rules and a test draft. Topic, language and timing confounds prevent causal claims about style.
  • Editorial plan: use the stated goals and available sources. Offer a schedule as a proposal; do not create an automation merely because a plan contains dates.
  • Authorized publication: verify target account, reviewed text/media, scope, publisher capabilities and existing delivery state before writing. User authorization already given need not be requested again.
  • Analytics or troubleshooting: inspect the requested account and time window, distinguish missing data from zero, and report what evidence supports the diagnosis.

Minimal context / 最少输入

Use context already supplied: audience, desired action, channel, language, source and deliverable. Ask only for missing information that changes the next action. A single source-linked note is enough for a draft; do not require a large material library or a fixed interview sequence.

For recurring operations also record the account handle/ID, timezone, approved cadence and current integration. Never infer credentials or publisher from the language of a post. Secrets belong in the configured credential store, not in drafts or status files.

Draft workflow / 起草

  1. Make a compact evidence card: source/excerpt, author, date, claim, verification status and allowed use. Private transcripts can calibrate voice but do not authorize public disclosure of their contents. Keep quoted authors separate from the poster.
  2. Select one reader problem and one useful takeaway. Preserve meaning before style. Numbers need a traceable source and scope; one authoritative primary source can suffice. Do not combine estimates with incompatible definitions.
  3. Choose format according to the user's request. Single post for one compact point; thread when the argument needs sequential parts. For a requested thread, deliver a thread draft even if publishing capability is unknown, marking delivery checks pending. No universal post length, thread count, daily frequency, opening formula or prohibited topic.
  4. Adapt to the channel without inventing first-person experience. Choose a CTA only when it serves the reader. If a promised link/comment is not yet deliverable, put the essential information in the draft itself.
  5. Deliver the draft plus source notes, uncertain claims, and any media outline. Do not mistake an outline for a rendered asset. State whether author review is still pending.

Read three-platform workflow only for a multi-platform publishing job. A single-channel draft does not require three variants, an image or a campaign.

Publication and recovery / 发布与恢复

Within authorized scope, check current platform documentation and the configured connector rather than copying historical API endpoints, credential fields or asset syntax. Maintain a capability table for root post, reply/thread, long text and media: verified / unsupported / unknown. Account subscription capabilities may differ from the publisher's capabilities.

Before a write: verify account identity, final text, attachment accessibility, source permission, timezone/schedule and duplicate state. A local file or successful upload is not proof of a published attachment. Public hosting is a separate disclosure and must be within the authorized scope.

After a write: record the returned ID, read its state, and verify the final public link, text and intended media. Distinguish draft, scheduled, sending, sent, failed and unknown. A log is a record, not the sole authority: reconcile it with platform readback. Never retry an ambiguous write blindly.

For ongoing operations, use an existing ledger or create a small task-local ledger as needed:

Source IDAccountReviewed copy/versionTarget time + timezonePublisher IDStatePublic URLLast verifiedNext action
[source][account][version][time][ID][state][URL][time][action]

Record only credential references, never secrets. A missing ledger does not establish that nothing was published. Check the platform before any recovery action. Recurring monitoring and notifications require a user request; do not auto-create them or send external alerts.

Measure and learn / 复盘

Record per-post date, format, topic, language, role, views/impressions if visible, interactions by type, link clicks and downstream outcomes. Define each rate's denominator and time window. If historical follower count is missing, label any current-follower comparison as an approximation. Referral sessions, impressions and follower changes are not interchangeable and do not by themselves prove ROI.

Compare like-for-like posts where possible. Keep alternative explanations such as topic, timing, audience and paid distribution explicit. Report small-sample uncertainty; propose the next bounded experiment instead of enforcing a successful-looking historical recipe.

Useful handoffs

For source-to-multiple-channel adaptation use gr-social-distill; for deeper author evidence use gr-brand-voice-calibration; for a paid creator campaign use gingiris-kol-outreach or gingiris-ugc-matrix. These are optional specialists: if absent, finish the bounded requested draft/plan here and identify any additional work separately.

Install

npx skills add Gingiris-1031/gingiris-skills --skill gingiris-twitter-agent-ops

Completion means the requested draft/plan is usable, or the authorized delivery is verified. Do not report publication or business impact from a draft, API acceptance or self-rating.

<!-- gingiris-community-cta:start -->

加入出海社群与购买资料包 · Community & Resources

  • 加入出海社群: 微信添加 staywildgrow,留言「加入社群」即可。
  • 购买资料包: 微信添加 staywildgrow,咨询购买 Gingiris 出海资料包,价格 599 元人民币。
  • Join the global growth community: Add staywildgrow on WeChat and send “加入社群” (join the community).
  • Get the resource pack: Add staywildgrow on WeChat to purchase the Gingiris global growth resource pack for CNY 599.

社群与资料包为可选入口,不影响本开源 Skill 的安装和使用。

The community and paid resource pack are optional; neither is required to install or use this open-source skill.

<!-- gingiris-community-cta:end -->

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.