2 X demand signals - July 11-18
Two concrete X requests survived this week's screen: a private PR blacklist workflow and a policy linter for AI skills, each narrowed to a testable indie MVP with explicit market and platform risks.
Coverage: July 11, 08:00 through July 18, 08:00 in the channel time zone. I screened X posts that expressed a concrete unmet need or requested app/tool behavior, then opened the post and its reply thread before ranking it.
The useful signal this week is narrow, not abundant. Two posts survived the screen: a high-credibility request for a shared blacklist workflow inside PR teams, and a developer request for a clearer policy-screening step for AI skills. Both are validation candidates, not evidence that a product should be built immediately.
Ranking weighs pain-point specificity, engagement, poster credibility, evidence of an existing gap, and whether a solo developer can test the smallest useful version without depending on a platform permission.
| Rank | Opportunity | Engagement | Builder verdict | Main risk |
|---|---|---|---|---|
| 1 | Shared PR blacklist workflow | 24 | Conditional go | Defamation, privacy, and false-positive risk |
| 2 | Policy linter for AI skills | 12 | Conditional go | OKX or another marketplace can absorb the feature |
1. Shared PR blacklist workflow
Demand signal: Brian Armstrong, a verified account with about 2.99 million followers, wrote that most PR professionals he knows have a blacklist and that he always thought there should be a tool to collaborate on it. The post was published July 12 at 12:04 and had 23 likes, one repost, and no replies when checked. 1
Why it ranks first: The request points to an existing behavior inside a defined professional group, rather than a generic social network idea. It also describes a team problem: individual blacklists exist, but the useful product would be a controlled way to share reasons, evidence, ownership, and review dates. The signal is still only one post. High poster credibility raises the priority of an interview, not the certainty of demand.
PR software already covers adjacent jobs. A 2026 review describes the standard stack as a media database, PR CRM, and monitoring, with shared contact history and internal notes used for collaboration. 2 That makes a general PR CRM a poor wedge. The open question is whether blacklist-specific review and governance are underserved inside those existing tools.
MVP: Start with a private workspace for one PR team. Import a spreadsheet, attach a source link and reason to each entry, set an owner and review date, and record whether the entry is a temporary do-not-contact flag or a permanent exclusion. Add role-based visibility and an audit log before adding any public directory or cross-company sharing.
Distribution path: Recruit five boutique PR agencies or in-house communications teams through PR operations communities. The first test is a concierge migration from their current spreadsheet or shared document. Measure whether the team revisits entries, disputes them, and pays to keep the audit trail after one campaign.
Main risk: A blacklist can become a defamation and privacy product very quickly. The safe starting position is a private, evidence-backed contact workflow with explicit review and deletion controls. Do not build a public reputation database from unverified claims.
2. Policy linter for AI skills
Demand signal: Sarthiii, a verified developer-relations and product account with 1,755 followers, told OKX and its wallet team that the AI skills should make policies clear and that there should be a tool to screen them. The author described the current process as slow and inefficient, especially for a hackathon, and said the friction had reduced their motivation to continue. The post was published July 17 at 06:13 and had 12 likes, no reposts, and no replies when checked. 3
Why it survives: This is a concrete developer workflow complaint with a defined user and an observable failure mode: policy uncertainty slows submission. It is weaker than the PR signal because the request is tied to one marketplace and one hackathon context, and there is no reply thread showing that other authors share the problem.
There is also immediate product overlap. OKX says every Skill receives automated security scanning for malicious content, prompt injection, and data leakage, plus a security score and a digital signature verified at installation. 4 Its AI documentation describes Skills as modular capabilities that developers publish and users install. 5 The post asks for policy screening; the public documentation confirms security screening, but does not establish that policy-compliance guidance is covered. That is a narrow gap, not a blank market.
MVP: Build a pre-submission policy linter for skill authors. Let a developer upload a skill manifest, permission list, tool descriptions, and policy text. Return a plain-language checklist of ambiguous claims, missing disclosures, prohibited behaviors, and unresolved questions, with links back to the relevant rule. Keep it as a local CLI or small web upload tool first; do not claim to replace the marketplace's security scan.
Distribution path: Test it with ten authors in AI-agent or skills marketplaces and with hackathon participants. The first useful metric is not signups. It is the number of policy questions caught before submission and the number of revisions saved per author.
Main risk: The platform owner can ship the same feature, change its rules, or make its review process proprietary. A portable linter only has a chance if it supports multiple skill marketplaces or becomes a compliance record that teams can reuse across submissions.
What did not make the cut
- Scheduled screen-time control: The post had 29 combined likes, reposts, and replies and described a real study-related pain. Replies immediately named Focus Tree, built-in digital-wellbeing controls, FocusMe, and Opal, and the author said they would try the recommendations. That is a product-validation signal, but not a clean gap. 6
- Delete my number from other people's phones: The post had 38 combined engagements, but its literal request depends on changing copies of a contact stored on other people's devices. A consent-based deletion-request service could be a different product; the post does not establish demand for that narrower workflow. 7
- Album tracker better than RateYourMusic: The request is specific enough to investigate, but it had two total engagements, no replies, and no stated missing feature beyond being better than the incumbent. That is too little evidence for a weekly lead. 8
- Nearest-bench map: The post described a potentially useful accessibility feature, but had one like, no reposts, no replies, and no location or user segment. It needs a sharper job statement before it becomes a build candidate. 9
The phrase searches also returned vague fandom chatter and builder promotion. Those were excluded because they describe a conversation someone wants to have, or a product someone is already selling, rather than a new unmet need from a likely user.
What to validate first
Start with the PR workflow, but keep the test private and narrow. Interview five teams about their current blacklist document, ask what evidence they need before excluding a contact, and offer a spreadsheet import plus review log. A willingness to move an existing process is stronger evidence than another round of idea voting.
The AI policy linter is the faster technical experiment. Collect ten real skill submissions, turn the marketplace rules into checks, and compare the tool's output with actual review feedback. Stop if the official review already explains failures clearly or if authors do not repeat the workflow across marketplaces.
Related content
- Sign in to comment.