Agentkit SEO X Twitter

AgentKit SEO X/Twitter

Overview

Use the X/Twitter hub to improve profile clarity and posting structure while avoiding overclaims about undocumented live ranking systems.

Reference selection

• Bio, profile, pinned post, media, thread structure: references/profile-and-posts.md
• Ready-to-post display name, bio, pinned post, post, thread, Alt Text: references/section-recipes.md
• Growth strategy, replies, ranking-signal explanations: references/engagement-and-ranking.md
• Premium-specific advice and confidence labeling: references/premium-and-confidence.md
• Existing-account audit or maintenance: references/account-audit-and-maintenance.md

Token discipline

• Do not analyze long posting history unless the user asks for an audit.
• For profile copy, inspect bio, pinned post, proof links, and up to 3-5 recent posts first.
• Read Premium or ranking references only when the user asks about those topics.
• Prefer pasted profile text, public profile fields, pinned post, links, and a small recent-post sample before asking for analytics exports or screenshots.
• Keep source ledgers compact: list input groups, not every post unless the post itself is discussed.
• Name next inspection if bounded.

Depth contract

Use the smallest honest audit depth:

• Quick scan: display name, bio, link path, pinned post, and obvious niche/proof gaps.
• Default audit: quick scan plus up to 10 recent posts, media/Alt Text when visible, proof-link alignment, and posting capacity assumptions.
• Deep audit: last 20-30 posts, reply behavior, topic drift, analytics screenshots, Premium capabilities, and cross-platform proof consistency.

Default to Default audit for broad account or profile requests. Offer Deep audit as an optional next step when the current answer would benefit from more evidence. Do not choose Deep audit silently unless the user asks for a full account audit, content system, analytics review, or recent-post history diagnosis.

Intake workflow

• If the user provides an X/Twitter URL or handle, inspect publicly accessible profile material and recent posts when tools allow it.
• If public access is blocked, stale, or incomplete, ask for screenshots, pasted bio and pinned post, recent post examples, analytics summaries, or a local text file export.
• Ask for the target audience, niche, posting capacity, proof links, and topics the user can credibly discuss before building a posting strategy.
• If the account strategy depends on professional facts or cross-platform consistency, recommend creating or updating the agent context file first.

Rules

• Treat Phoenix, Grok, and related architecture clues as design signals, not as a complete live-production contract.
• Verify current official X product documentation before giving paid-tier, Premium, post-length, media-length, monetization, or account-capability advice. If verification is unavailable, label the guidance as historical or inferred.
• Separate facts verified from public account material, facts supplied by the user's context material, and recommendations inferred from those facts.
• Do not promise ranking outcomes.
• Do not infer private analytics, Premium status, shadowban status, or ranking treatment from incomplete public views.
• Keep recommendations aligned with the user's actual niche, expertise, and posting capacity.
• Distinguish official product features from empirical tactics.
• Keep profile positioning, pinned assets, posting topics, and linked external proof aligned around one clear niche.
• When profile proof, audience, or posting history is missing, ask for it before inventing claims or forcing a content strategy.

Response shape

Return only requested-relevant sections. Do not add cadence or engagement strategy unless requested or clearly necessary. For audits, return:

1. public inputs inspected and any blocked inputs
2. profile and content-positioning diagnosis
3. ready-to-paste bio, pinned post, thread, or post drafts
4. cadence and engagement recommendations sized to the user's capacity
5. missing inputs needed for a stronger second pass

For audits, use concise labels such as Verified, Official feature, Historical/open-source inference, Empirical tactic, From context, Inference, and Inaccessible when a claim could otherwise be ambiguous. When the audit is intentionally bounded, include a one-line Depth note that says what profile/post scope was inspected, what was not inspected, and what deeper inspection would add.