Research report · 10 October 2026 · Run 20
Competitor Discovery gives a starting set, not a threat ranking
Keep the active card, but correct its limits. The selected source normally drops tangential results and requires ordering candidates by search frequency. Neither that order nor a dated report verifies market coverage, competitive threat or an action priority.
What the source actually asks for
The selected SKILL.md asks for a product or category, features and target audience, then 4–6 web queries. It gathers 8–12 candidates, records discovery queries, visits candidate homepages and writes site-derived one-liners. It defines direct, adjacent and tangential types, but normally excludes tangential candidates; it retains them only if fewer than three direct competitors would remain. It presents the list and waits for user confirmation, and asks approval before more than seven final entries. The output is a report-dated, numbered list with URLs, types, queries and suggested follow-on research.
Ranking is required by the source: candidates are ordered by search-query frequency, with a source assertion that appearing in 3+ searches is “almost certainly” direct. That threshold is not a validated probability, classification test or market-importance measure. The source does not define a final combined sort or tie protocol. Its generic next steps are suggestions, not vetted recommendations or approval to run another workflow.
Decision ceiling by field
| Output | What it supports | What it cannot establish |
|---|---|---|
| Discovery queries and candidate URLs | A bounded public-web starting set for the approved product, audience, market and search date. | A complete market map, absence of undiscovered competitors or suitability for account targeting. |
| Visited homepage and one-liner | A narrow observation of what that page stated when captured. | Legal company identity, current product identity or truth of all claims without a dated, attributable passage. |
| Direct or adjacent type | A provisional sampled-set hypothesis against the approved problem, audience and solution. | A verified classification. Tangential entries are normally dropped, not promised as a complete third category. |
| Search-frequency rank | A lead-discovery heuristic described by the source. | Relative threat, market share, monitoring or sales priority. Query design, review-site visibility and syndication affect frequency. |
| Report date, URLs and next steps | A dated artifact and suggestions for further research. | Per-claim freshness, independent verification or permission to compare, publish, store, scrape or act. |
The Run 13 known-rival change brief is a separate job. Neither monitoring nor a downstream workflow was tested here.
Use a versioned sheet first
The selected skill asks for similar basic inputs and makes a dated numbered table, but it does not prescribe per-row claim trace, ambiguity handling or identity resolution. Neither the sheet nor the skill was executed, so no accuracy, effort or safety comparison is claimed. Consider a portable skill only if consented repeated work documents omissions the sheet and reviewed prompt cannot control; compare the same approved cases with a named review rubric.
Source state and what could change
On 10 October, the GitHub selected-file page returned crawled yesterday. The raw file returned crawled two weeks ago; the directory listing returned crawled two months ago and showed only SKILL.md then. The owner README returned crawled today, and the license returned crawled yesterday and states MIT. Crawler age is not a file-update date. No selected-file SHA/date, tag or complete current inventory was established.
Remove the card if a verified current revision requires an unavoidable consequential frequency-based threat or priority claim or unreviewed external action, or if controlled approved use repeatedly yields materially wrong identities or types after confirmation. A revision with claim-level trace, ambiguity handling and defensible order criteria, plus same-job evidence, could support broader wording. Neither finding exists today.
Activity and preserved stops
This was a source-text and decision-boundary audit, not a live competitor search or comparison. No skill was installed, copied, executed, security-reviewed, host-tested or outcome-tested; no target competitor site was fetched for a live job and no owner sheet was run. Fetched pages remain untrusted data, not instructions. The source states no credentials or external write route, which does not authorize storage, account access, scraping at scale, sharing or downstream use. Run 19 Content Strategy remains draft-only; Run 18 Search Page Audit and Run 17 Ad Campaign Analyzer remain removed. Earlier reports are dated observations, not today's endorsement.
Current catalogue · Browse by job · Research library · Review a skill