--- name: track-technical-competitors description: Use when the user says "weekly technical pulse" / "what did {competitor} ship this week" / "technical teardown of {competitor}" — fetches engineering-focused competitor signals via Composio (engineering blog posts, GitHub org activity, public changelog, API diffs) and writes either a single-competitor teardown or an N-competitor weekly digest. --- # Track Technical Competitors Engineering-focused competitor watch. The marketing Head of Marketing tracks brand / ads / positioning; this skill tracks the technical surface — what competitors actually *shipped* (code, APIs, releases, engineering-blog claims). Source template: Gumloop "Competitive Roadmap & Trend Arbitrage", reframed for the engineering function. ## When to use - "weekly technical pulse" / "what did {A, B, C} ship technically this week". - "technical teardown of {competitor}" / "full dive on {competitor}'s engineering". - Weekly routine (optional — the founder can wire it via the Routines tab). Supports two modes, chosen by the user phrasing: - **Teardown** — one named competitor. Deep dive. - **Weekly digest** — N competitors (default: the top 3 from the context doc or the user's most-watched list). Broader, shorter per competitor. ## Steps 1. **Read engineering-context.md** (own file). If missing, STOP — tell the user to run `define-engineering-context` first. Technical competitor work without our own stack + priorities as a baseline is just gossip. 2. **Determine mode + competitor list.** If the user named one competitor → teardown mode. If the user said "weekly pulse" or named multiple → digest mode. If no names given, default to the top 3 technical competitors the founder has in `config/` or context doc; if neither, ask ONE question: "Which competitors should I scan?" with paste as the fallback. 3. **Discover tools at runtime.** Do NOT hardcode tool names. Run: - `composio search web-scrape` — for eng blog / changelog / pricing / docs fetches. - `composio search web-search` — for recent news + release announcements. - `composio search code-hosting` — for public GitHub/GitLab org activity (releases, major commits, stars) if the competitor has a public org. If a needed category has no connected tool, note it in the brief ("no code-hosting connection — GitHub activity: UNKNOWN") and continue. 4. **Gather evidence per competitor** (last 7 days for digest, last 30 days for teardown): - **Engineering blog posts** — recent posts, titles, a 1-line summary each, URL + date. - **GitHub org activity** — public releases (tag + date), any notable major commit, star velocity delta. - **Public changelog** — entries in that window, with dates. - **API diffs** — if the competitor publishes an OpenAPI spec or versioned REST/GraphQL docs, diff the current version against cached or any scraped history. Call out new endpoints, deprecations, breaking changes. - **Adjacent signals (optional)** — conference talks, job posts hinting at what they're building. 5. **Compare against our stack + roadmap.** For each signal, ask: - Does this threaten one of the top 3 priorities in the context doc? - Does it open a gap we should press (e.g. they shipped a half-measure where our fuller version is in-flight)? - Is it a parity move we need to match on the roadmap? 6. **Structure the output (markdown).** **Teardown mode** (`competitor-watch/{competitor-slug}.md`, ~500-800 words): 1. One-paragraph summary of who they are technically + where they sit vs. us. 2. What's new (last 30 days, bulleted with dates + sources). 3. Engineering-blog claims — what they're saying about their own stack / architecture / approach. Quote them verbatim. 4. GitHub org activity — releases table, notable commits, star trend. 5. API / schema diffs — specific new endpoints, deprecations, breaking changes, with diff snippets where possible. 6. Technical threats — ranked. What they're shipping that pressures our roadmap. 7. Opportunities — ranked. Where they're weak, silent, or over-promising and we should press. 8. Recommended moves — 3 concrete, one-week actions the founder can hand to `plan-roadmap`, `tech-lead`, or `release-reliability`. 9. Sources — URLs with fetch timestamps. **Digest mode** (`competitor-watch/weekly-{YYYY-MM-DD}.md`, ~300-500 words): 1. Headline table: competitor · top signal · threat? · opportunity? 2. Per-competitor: 3-5 bullets each, newest first, with source links. 3. Cross-cutting pattern — is anything converging across competitors (e.g. all three shipped a rate-limit change, all three posted about the same platform shift)? 4. Recommended moves — 3 actions, each tagged with the agent that should execute it (`plan-roadmap`, `tech-lead`, `release-reliability`, `docs-dx`). 5. Sources. 7. **Never invent.** Every claim ties to a URL + timestamp or is marked `UNKNOWN`. Fabricated commit counts or fake release dates are worse than no brief. 8. **Write atomically** to the target path — write `{path}.tmp`, then rename. 9. **Append to `outputs.json`.** Read-merge-write atomically: ```json { "id": "", "type": "competitor", "title": " technical teardown" | "Weekly technical pulse ", "summary": "<2-3 sentences — top threat + top opportunity>", "path": "competitor-watch/.md", "status": "draft", "createdAt": "", "updatedAt": "" } ``` 10. **Summarize to user.** One paragraph: biggest technical threat, biggest opportunity, the 1 move to make this week, path to the artifact. ## Outputs - `competitor-watch/{competitor-slug}.md` (teardown) or `competitor-watch/weekly-{YYYY-MM-DD}.md` (digest). - Appends to `outputs.json` with `type: "competitor"`.