Find My Bottlenecks

Find My Bottlenecks

When to use

• User asks "what's stuck," "what's blocking progress," "why aren't we moving on X."
• Most recent weekly review (from this agent) repeats risk or ask from prior one.
• Goal flipped to off-track and linked initiative also slipped.

Connections I need

I run external work through Composio. Before this skill runs I check the categories below are linked. Missing → I name the category, ask you to connect it from the Integrations tab, stop.

• Project tracker (Linear, Notion, Asana) - Optional. Surfaces stalled initiatives and ticket-level blockers; I work without it but with less signal.
• Team chat (Slack) - Optional. Lets me catch repeating asks across threads.

This skill works without any connection - it leans mostly on what's already in your saved work. I never block here.

Information I need

I read your operations context first. For every required field that's missing I ask ONE plain-language question (best modality: connected app > file drop > URL > paste) and wait.

• Active priorities - Required. Why I need it: a bottleneck only matters if it blocks something you're pushing on. If missing I ask: "What are the 2 to 3 things the company is pushing on this quarter?"
• Key contacts - Required. Why I need it: I propose owners to unblock; without contacts I'd be guessing names. If missing I ask: "Who unblocks what - engineering, sales, ops? Names plus how to reach them."
• Recent decisions, reviews, or goal snapshots - Optional. Why I need it: more saved work means stronger evidence. If you don't have it I keep going with TBD and lean on what's there.

Steps

<!-- houston-workflow:v1 -->

1. Read context/operations-context.md. If missing or empty, stop and ask user run set-up-my-ops-info first. Priorities and key-contacts anchor "proposed owner to unblock" logic.

2. Gather evidence from last 4 weeks (handle each source as "if present, use it; if missing, continue"):

   • reviews/ - last 4 weekly review files. Look for repeating risks / asks.
   • triage/ - last 4 inbox-triage files. Recurring "can-wait" threads from same person hint at delegation bottleneck.
   • decisions.json - any decision with status: "pending" older than 14 days → decision-latency bottleneck.
   • goal-history.json - any goal metric off-track across two or more consecutive snapshots → bottleneck candidate tied to linked initiative.
   • this agent anomalies.json - repeating open anomalies hint at data-or-process bottleneck.

3. Cluster recurring themes. Group evidence by shared owner, shared cross-team dependency, or shared goal. Bottleneck = cluster, not individual incident.

4. For each cluster, form hypothesis (1-2 sentences, never stated as certain):

   • "Hiring in engineering bottlenecked on founder's interview calendar - 3 initiatives waiting on same reviewer."
   • "Pricing changes blocked on pending decision from week of {date} - 2 launches staged behind it."
   • "Cross-agent data pulls duplicating work - both board pack and investor update asking for same retention query."

5. Propose owner to unblock. Read leadership / key contacts section of operating context. For cross-team bottlenecks, owner = whoever owns blocking resource (e.g. CTO for engineering-calendar constraint), not downstream exec. For solo founder, owner = founder - proposed unblock usually "carve time for {X}" or "delegate {Y}".

6. Quantify impact. List impactOnGoalIds (objectives blocked) and impactOnInitiativeSlugs (initiatives stalled). Keep citations tight - evidence strings reference real paths (review files, decision slugs, anomaly ids).

7. Dedupe against open bottlenecks. Read bottlenecks.json. If cluster matches existing open row (same proposed owner + overlapping impact set), update in place (add new evidence, refine hypothesis, refresh updatedAt). Do NOT create duplicate.

8. Sensitive-matter routing. If hypothesis names specific person as bottleneck (performance / capacity), do NOT land that language in bottlenecks.json. Generalize to role-and-process language ("engineering interview capacity") in index row. Flag specifics to CEO in chat only.

9. Write new / updated bottlenecks to bottlenecks.json (atomic). Each row: { slug, title, hypothesis, proposedOwner, impactOnGoalIds, impactOnInitiativeSlugs, status: "open", evidence, createdAt, updatedAt }.

10. Append to outputs.json with type: "bottleneck", status "ready" per new row.

11. Hand off in chat.

    {N} bottleneck(s) identified.
    
    1. **{title}**  -  proposed owner: {owner}.
       Hypothesis: {hypothesis}
       Blocks: {N} goal(s), {M} initiative(s).
       Evidence: {citations}
    
    2. ...
    
    Want me to draft a nudge to {proposed owner} for #1?
    (I'd hand that off to the `draft-a-message` skill.)