with one click
opendevbrowser-research
// Skill-guided, evidence-gated research workflow for provider-constrained public source gathering and auditable artifact review.
// Skill-guided, evidence-gated research workflow for provider-constrained public source gathering and auditable artifact review.
This skill should be used when the user asks to design or run OpenDevBrowser provider workflows, scraping pipelines, QA/debug automation, parity checks across modes, or resilient browser operations with codified scripts and artifacts.
This skill should be used when the user asks to design, redesign, audit, or implement frontend/UI work with OpenDevBrowser, screenshots, or the /canvas surface. It turns briefs and existing interfaces into contract-first, real-browser-validated design execution.
Deterministic motion and animation design guidance for OpenDevBrowser agents building web and mobile UI with real-browser temporal evidence.
Collects product metadata, copy, and optional visual assets into a local folder pack for video workflows.
This skill should be used when the user asks to "extract data from a page", "scrape tables", "collect paginated results", "parse list/card content", or "export structured web data" with OpenDevBrowser.
This skill should be used when the user asks to "test a form", "verify validation", "check submission behavior", "test multi-step forms", or "debug form errors" with OpenDevBrowser.
| name | opendevbrowser-research |
| description | Skill-guided, evidence-gated research workflow for provider-constrained public source gathering and auditable artifact review. |
| version | 2.1.0 |
Load this skill before research tasks. Use it to plan source families, gather provider-constrained evidence, review artifacts, and publish only claims that survive the evidence gate.
opendevbrowser research run is a low-level, best-effort primitive. It can collect and render provider results, but the skill owns source planning, blocker review, confidence, limitations, and final synthesis.
artifacts/research-workflows.mdassets/templates/compact.mdassets/templates/context.jsonassets/templates/report.mdscripts/run-research.shscripts/render-output.shscripts/write-artifacts.shscripts/validate-skill-assets.sh../opendevbrowser-best-practices/artifacts/browser-agent-known-issues-matrix.md./skills/opendevbrowser-research/scripts/validate-skill-assets.sh
./skills/opendevbrowser-research/scripts/run-research.sh "browser automation" 30 context "web,community"
./skills/opendevbrowser-research/scripts/render-output.sh "browser automation" compact
screencast-start / screencast-stop) when claim review needs temporal UI evidence or parity proof from browser-backed runs.--challenge-automation-mode off|browser|browser_with_helper for bounded browser-scoped computer use in research workflows; it is not a desktop agent.--days or --from/--to).web, community, social, shopping, or a deliberate combination such as web,community.auto as a source-family selector, not a reliability guarantee.Review artifacts before publishing claims. A successful command exit or rendered report is not enough.
Preserved artifact files:
summary.mdreport.mdrecords.jsoncontext.jsonmeta.jsonbundle-manifest.jsonRequired review:
records.json for fetched source records, timestamps, providers, extraction quality, and blockers.context.json for source ledger, evidence gaps, unsupported claims, staleness checks, and search-engine provenance when used.meta.json for provider limits, warnings, no-evidence failures, cookie diagnostics, challenge/auth/token failures, and artifact generation details.--browser-mode extension for an existing signed-in relay session, --use-cookies only when legitimate cookies are available, and --challenge-automation-mode browser_with_helper for browser-scoped assistance.report.md and summary.md only after confirming claims map back to accepted evidence.Research is iterative. Do not stop at the first provider page or SERP shell.
/r/.../comments/... result URLs over /login, /search, /account, /submit, or root navigation.--browser-mode extension, --use-cookies, or --challenge-automation-mode browser_with_helper only when the selected evidence page itself needs authorized browser recovery and the user has legitimate access.This lane is optional, skill-guided, provider-constrained, and discovery-only. It can improve breadth, but it is not a reliable default and does not replace opendevbrowser research run or the evidence gate.
../opendevbrowser-best-practices/SKILL.md ("Parallel Operations").managed, extension, and cdpConnect where browser-backed evidence capture is used.Matrix source: ../opendevbrowser-best-practices/artifacts/browser-agent-known-issues-matrix.md
ISSUE-06: upstream throttling and retry disciplineISSUE-09: pagination/result drift across sourcesISSUE-10: locale/currency normalization consistencyISSUE-12: stale evidence and claim traceabilitydays or from/to).opendevbrowser research run as a low-level best-effort primitive.records.json, context.json, and meta.json before trusting report.md.meta.json shows auth, token, challenge, or cookie-gated providers, make the next run skill-first only after the candidate queue has no relevant public destination evidence left: use the existing signed-in browser session when authorized, cookies only when legitimate cookies are available, and browser-scoped challenge assistance only for that browser session.opendevbrowser research run --topic "<topic>" --days 30 --sources web,community --mode context
opendevbrowser research run --topic "<topic>" --sources web --mode json
opendevbrowser research run --topic "<topic>" --sources web,shopping --mode md
auto and all are selector values in the current source-family contract, not promises of reliable coverage.--source-selection shopping or explicit --sources ...shopping... to include shopping only when commercial intent is explicit.--mode path with scripts/write-artifacts.sh when you need replayable handoff bundles.../opendevbrowser-best-practices/SKILL.md.