v0.14.1 candidate — Adoption onboarding

What changed

• Added state-aware paths for already-installed users, managed installers, and the audited Release ZIP fallback.
• Added bilingual 60-second first-use guidance with a real parseable research request and local workspace return command.
• Added lightweight bilingual animated workflow images with static SVG accessibility fallbacks.
• Fixed the documented English product-demand example so it resolves without an avoidable clarification.
• Added regression checks for onboarding routes, visual assets, and the three-user acceptance record structure.

Candidate boundary

External uncoached onboarding validation by three users has been deferred. This release improves and documents onboarding, but it does not claim that unfamiliar users have already completed the full path without assistance.

The Skill does not predict virality, traffic, demand, or revenue, and it does not package cookies, credentials, browser sessions, or live platform data.

Install

npx skills add datou202307-design/trend-opportunity-radar -g

v0.14.0 candidate — Repeatable local research workspace

What is new

• Local research workspace: reopen standardized runs from one private HTML view, separate unfinished studies, completed reports, due collection recommendations, and monitoring cycles, then hand an exact continuation prompt back to an Agent.
• Bilingual five-scenario case gallery: understand the output before connecting a platform through synthetic examples for business opportunity discovery, brand sentiment, competitor-user research, content opportunities, and product-demand validation.
• Clearer Chinese first screen: improved Chinese gallery line breaks, spacing, and mobile readability without removing useful context.

Trust and privacy boundaries

• The workspace reads run manifests, formal reports, and monitor state only. It does not copy raw platform content, upload local research, or expose absolute paths.
• A recommended monitoring cadence is not presented as a scheduled task unless an external scheduler actually succeeded.
• Legacy reports without a standardized run-manifest.json are not silently guessed or migrated.
• The case gallery is synthetic and does not claim customer outcomes or real market demand.

Validation

• 317 automated tests passed.
• Open-source audit and Skill structure validation passed.
• The workspace passed a read-only test against a real standardized local run tree, deterministic rebuild checks, link checks, and desktop/mobile loopback rendering checks.
• The installable archive was independently reproduced with one trend-opportunity-radar/ root and no tests, caches, credentials, browser sessions, local outputs, or live platform data.

Download the ZIP asset below, verify it with the attached SHA-256 file, extract it, and copy the single trend-opportunity-radar/ folder into your Agent Skill directory.

v0.13.0 candidate — First trustworthy result

This candidate reduces the distance from discovering the Skill to seeing a trustworthy first result. Users can now inspect a complete bilingual report without connecting a platform, then start a real study with only a topic and platform.

Highlights

• no-login synthetic Demo in English or Chinese, using the same report generator as a live study
• prominent synthetic-data markers in HTML, Markdown, JSON, and the Demo manifest
• deterministic two-input init flow for topic plus platform, feeding the existing preflight and evidence gates
• Windows PowerShell and Ubuntu/macOS first-run commands
• installable single-folder Skill ZIP generated from a strict runtime-file allowlist
• SHA-256 checksum and per-file manifest attached to the Release
• deterministic, idempotent first-success artifacts that refuse to overwrite changed files

Validation

• 305 automated tests passed
• official Skill structure validation passed
• open-source release audit passed
• Chinese and English Demo reports passed desktop and 375 px mobile browser QA
• no horizontal overflow or browser-console errors were observed
• package QA confirmed one Skill root and 104 runtime files, excluding tests and local state

Candidate boundaries

• Demo content is synthetic and is not evidence of live platform demand
• init stores no platform data, credentials, cookies, or browser sessions
• live research still requires lawful read-only platform access and successful per-run preflight
• comments, observed heat, and evidence confidence remain distinct signals
• the Skill does not predict virality, traffic, demand, revenue, or future performance
• compatible monitoring remains a candidate workflow; the first real three-day forward comparison is still pending

v0.12.0 candidate — Resumable evidence execution

This candidate makes Trend Opportunity Radar substantially more reliable across agents and models by turning the research instructions into a resumable, evidence-bound execution workflow.

Highlights

• unified start, doctor, and resume entry point with one manifest state and one explicit next action
• deterministic continuation through normalization, cluster audit, scoring, route proof, and three-format reporting
• immutable stage receipts that bind inputs and outputs and prevent silent step skipping
• frozen search, detail, comment, and media route proof before a live report can be delivered
• safer query recovery after semantic review, with bounded detail backfill and no duplicate sample inflation
• comment prominence and recurring comment-demand analysis while keeping comments separate from post trend volume
• compatible monitoring commands: create, append, and compare, with JSON, Markdown, and local HTML output
• clearer bilingual setup guidance, research-basis counts, evidence boundaries, and plain-language decision reporting

Validation

• 294 automated tests passed
• official Skill structure validation passed
• open-source release audit passed
• GitHub Release audit passed
• real X plus DokoBot interruption and resume test preserved completed-query counts without repeating the first query
• synthetic four-snapshot monitoring replay passed
• monitoring HTML passed desktop and 390 px mobile browser QA without horizontal overflow or console warnings

Candidate boundaries

• compatible monitoring is available as a candidate workflow, but the first real three-day forward comparison is still pending
• cadence recommendations do not create or imply an operating-system, Codex, or third-party scheduled task
• observed heat and evidence confidence remain separate; snapshot differences do not prove demand growth, causality, traffic, revenue, or future performance
• live platform reading still requires a lawful, explicitly authorized adapter and a successful per-run read-only preflight
• no cookies, tokens, browser profiles, customer data, live captures, internal brands, or private research material are packaged

v0.11.0 candidate - Facebook research and platform-native reports

This release adds bounded Facebook topic research and platform-native report interpretation for Facebook and Instagram.

Highlights:

• Facebook Posts topic-research Beta for authorized signed-in browser sessions, with bounded read-only collection, human-paced navigation, detail verification, and limited visible-comment review
• platform-native report interpretation: Facebook emphasizes public discussions, experiences, objections, and group/page context; Instagram emphasizes format, captions, visual/media evidence, and then comments
• strengthened Instagram Hashtag reporting with clearer content-supply versus user-demand boundaries
• English and Chinese GitHub platform tables and hero banners now include Facebook Beta

Validation:

• 237 automated tests passed
• open-source release audit and Skill structure validation passed
• real Facebook and Instagram HTML reports passed desktop and 390 x 844 browser QA

Boundaries:

• Facebook remains Beta and requires explicit opt-in plus per-run preflight
• Instagram visible Hashtag counts describe content supply, not search demand or trend growth
• No cookies, credentials, browser profiles, customer data, internal brand assets, live captures, or private data are packaged
• This release does not predict virality, traffic, revenue, or future trend direction

v0.10.0 candidate — Validated Instagram topic research

This release validates Instagram Hashtag topic research as a bounded, evidence-backed workflow.

Highlights:

• three frozen query layers with two paced reads per Hashtag
• deterministic OpenCLI browser capture, canonical-link union, and sequential detail reads
• multi-snapshot merge with per-layer and global sampling gates
• full semantic review plus bounded representative-comment review
• local HTML, Markdown, and JSON reports with visible collection counts and comment-derived requirements
• updated bilingual GitHub banners and documentation

Real acceptance run:

• 3 queries, 72 observed links, 30 unique reviewable posts, 18 details
• 28 relevant posts, 6 counter signals, 65/65 comments reviewed
• all sampling gates and browser QA passed

Boundaries:

• Instagram account research remains a separate pilot.
• No cookies, credentials, browser profiles, private data, or real captures are packaged.
• This release does not predict virality, search demand, traffic, revenue, or future trend direction.

Trend Opportunity Radar v0.9.0 Candidate

Trend Opportunity Radar v0.9.0 Candidate

This candidate expands the independent, brand-neutral Skill with a bounded TikTok topic-research Beta and a reusable video-evidence layer. It does not claim anonymous TikTok collection, unrestricted browser access, or predictive accuracy.

Highlights

• Adds explicitly enabled TikTok topic research for user-authorized, already logged-in Chrome sessions.
• Adds bounded OpenCLI search parsing, DokoBot detail enrichment, and a deterministic two-stage fallback for up to five visible top-level comments.
• Keeps displayed comment totals separate from captured comment bodies and treats comment enrichment as optional evidence.
• Adds a platform-neutral video-evidence contract for native captions, local ASR, key frames, and OCR, with provenance preserved for every channel.
• Adds human-paced, single-concurrency collection rules and runtime checks for optional video tooling.
• Improves adapter capability diagnostics, partial-result handling, query exhaustion rules, and collection-ledger reporting.
• Improves local HTML, Markdown, and JSON reports for video evidence and explicit evidence boundaries.
• Updates the English and Chinese GitHub presentation to show X, Xiaohongshu, YouTube, and TikTok Beta without overstating platform support.

Safety and scope

• No credentials, cookies, browser sessions, platform data, customer data, or internal brand assets are packaged.
• TikTok Beta requires explicit enablement and a lawful, user-authorized logged-in session.
• Collection remains read-only, bounded, non-concurrent, and subject to per-run capability checks.
• Douyin support and anonymous TikTok live research are not claimed.

Verification

• Full automated test suite: 157 tests passed on both Windows and Ubuntu CI-equivalent environments.
• Skill structure validation, open-source release audit, and source/install parity checks.
• Forward acceptance on a fresh TikTok topic: 60 observed and unique signals, 53 relevant signals, 12 opened details, 25 counterevidence items, and 5 reviewed comments.
• Desktop and mobile visual QA for the generated local report, plus rendered English and Chinese GitHub hero review.

This is a candidate prerelease intended for cross-environment evaluation before a stable release.

Trend Opportunity Radar v0.8.0 Candidate

Highlights

• Adds YouTube as a validated third platform alongside X and Xiaohongshu.
• Keeps the same five decision profiles: business opportunities, brand sentiment, competitor users, content opportunities, and product-demand validation.
• Adds bounded YouTube search and video-detail collection through the optional OpenCLI adapter.
• Preserves representative comments as separately reviewed qualitative evidence without counting them as trend samples.
• Adds a portable cross-platform comparison generator while keeping each platform's samples, heat, and confidence separate.
• Updates the English and Simplified Chinese README files and hero banners for the three-platform workflow.

Evidence and access boundaries

• Every live run must pass its own read-only capability probe; an installed CLI alone is not proof that the current platform session is usable.
• YouTube comments are separately requested and capped at 10 representative items per eligible video.
• Transcripts are opened only when needed to verify a video claim and are never guaranteed.
• Platform-specific engagement weights remain transparent candidate settings for within-platform ordering only; they are not calibrated causal values.
• The Skill does not package credentials, cookies, browser sessions, captured platform data, or private customer material.

Validation

• 116 automated tests passed.
• Platform adapter and five Decision Profile registries validated.
• Bilingual SVG banners passed XML and rendered-layout review.
• Open-source release audit and GitHub Actions release checks passed.

This remains a candidate release. It supports constrained, evidence-backed research and does not predict virality, traffic, demand, or revenue.

Trend Opportunity Radar v0.7.0 Candidate

Trend Opportunity Radar v0.7.0 Candidate

This candidate upgrades the Skill from a single opportunity-report workflow into a constrained research capability with five decision profiles:

• business opportunity discovery;
• brand sentiment monitoring;
• competitor-user research;
• content opportunity discovery;
• product-demand validation.

Highlights

• Short natural-language requests are compiled into a versioned research context.
• X and Xiaohongshu use the same adapter-neutral collection contract with validated OpenCLI/DokoBot selection.
• Sampling, semantic review, counterevidence, clustering, scoring, and provenance remain shared across decision profiles.
• Reports adapt their questions, evidence roles, actions, and information hierarchy to the user's decision goal.
• Repeated capture attempts are preserved as immutable files instead of overwriting earlier evidence.
• Human-facing HTML and Markdown keep internal collection gates out of the main reading path while JSON retains the full audit trail.

Validation

• 5 profiles × 2 platforms completed real end-to-end baseline acceptance.
• 10/10 existing real snapshots passed current-code replay.
• 20/20 diverse-topic snapshot replays satisfied the release matrix.
• Release audit: ready, gaps 0.
• 99 unit, contract, and regression tests passed in source, public-repository, and installed copies.
• Open-source leakage audit and Skill structure validation passed.

Boundaries

This remains a candidate research workflow. It does not predict virality, traffic, market size, paid demand, or revenue, and it does not compare raw heat scores across platforms. Live collection requires lawful user-authorized access to the target platform.

Trend Opportunity Radar v0.6.0 Candidate

Candidate release of the independent, brand-neutral Trend Opportunity Radar Skill.

Highlights

• Adds explicit quick, standard, and deep sampling contracts with auditable collection ledgers.
• Orchestrates bounded, read-only DokoBot collection for X and OpenCLI/DokoBot collection for Xiaohongshu where the required environment and authorized session are available.
• Adds deterministic query recovery, zero-result handling, detail backfill, semantic review, counterevidence, clustering audits, and evidence-quality gates.
• Separates observed heat from evidence confidence and keeps incomplete evidence from being presented as a software error or a prediction.
• Produces mutually consistent local HTML, Markdown, and JSON reports, with loopback-browser visual QA for HTML delivery.
• Improves reader-facing language, opportunity titles, evidence-boundary summaries, and optional follow-up monitoring guidance.
• Adds an automated open-source release audit and GitHub Actions validation workflow.

Validation

• 58 pipeline tests passed.
• Open-source release audit passed.
• Skill structure validation passed.
• The public repository contains synthetic fixtures only; no captured platform data, browser sessions, credentials, customer material, internal brands, or machine-specific run artifacts are packaged.

Boundaries

This is a trend-research workflow, not a viral-content, traffic, demand, or revenue prediction system. DokoBot, OpenCLI, Chrome, X, and Xiaohongshu are optional third-party tools or platforms and are not bundled or endorsed. Users remain responsible for lawful access, platform terms, and account permissions.