agentleFS
sambersamber/developer-platform-skills/skills/stackoverflow-playbook/SKILL.md

stackoverflow-playbook

Skill9 starsChanged 23 days agoMITView on GitHub

Runs a developer-tools vendor's public presence on Stack Overflow - product-tag ownership, staff answering strategy, affiliation disclosure and self-promotion limits, reputation gates, the generative-AI content ban, backlog triage, and canonical self-answered Q&A that ranks in search. Use whenever the user mentions their product tag on Stack Overflow, staff answering there, unanswered questions in their tag, staff answers flagged as spam, or "is Stack Overflow still worth it for our product" - even if they only say "our tag is a mess". Not Stack Internal, ads or the Developer Survey. Do NOT use for getting your own docs cited by AI - use samber/developer-relations-skills@developer-docs-ai-visibility; for docs error pages use samber/developer-relations-skills@developer-troubleshooting-docs.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    One person, one account: No shared or agency accounts.

    line 30
  2. Distinctive:

    Hands off others' work: Never downvote, flag or retag competitors' posts.

    line 31

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Stack Overflow playbook

You run a vendor's presence on Stack Overflow: a reference archive that search engines rank and licensed AI assistants retrieve, not a live support queue.

Rules that get staff accounts flagged

The first four are published in the help center.

  • Disclose in every post naming your product or employer, including the question half of a self-answered pair. A profile bio does not count.
  • Answer the question, not the product. Some answers may be about your product, never all. Show the working solution; link-only answers are deleted.
  • No generative AI. Content "crafted, in part or in whole" by an LLM is banned: deletion with its reputation, warning, suspension on repeat.
  • One person, one account. No shared or agency accounts. Colleagues never vote on, accept or bounty each other's posts; serial votes are reversed automatically.
  • Hands off others' work. Never downvote, flag or retag competitors' posts. Never rewrite another author's solution; edits below 2,000 rep that change intent are rejected, and rewrites get rolled back. Post a new answer instead.
  • Never paste one answer on many questions; flag copies as duplicates. Posts are CC BY-SA 4.0, so reuse others' answers in docs with attribution.
RationalizationWhy it fails
"The profile says where they work."Disclosure belongs in the post.
"AI only drafted it; an engineer checked it.""In part" is covered. Debug with AI privately, write the post yourself.
"Ask the team to upvote our canonical."Coordinated votes are reversed and can suspend every account involved.

Interview

One question at a time; skip what the request already answers:

  1. Who answers, how many accounts are new, what reputation each?
  2. Tag state: questions, unanswered, highest-view threads? If unknown, run the Data Explorer query in the templates.
  3. Where do bugs, feature requests and account issues go today?
  4. Goal: deflect support, be cited in search and AI answers, or wind down? Anyone planning AI drafting?
  5. Hours per week, and the date a result must show by.

Decide the posture

Monthly new questions sit under 1% of the 2014 peak. The archive still ranks in search, is licensed to Google and OpenAI, and third-party studies find it among domains AI assistants cite for how-to queries. Restate "clear the backlog" as owning the high-view threads.

  • Curate (default): new questions still arrive monthly; goal is citation or deflection.
  • Maintain: under about 2 hours a week; canonicals and tag excerpt only.
  • Exit: almost no new questions; excerpt points to your channels, update top-viewed answers once, stop.

Engagement surfaces, ranked by value per hour

Efficiency: tag excerpt and wiki > self-answered canonical > new answer on a high-view outdated thread > duplicate flags > new-question rotation == Staging Ground review (one new question each, same cost) > bounty. A citation goal promotes canonicals, a deflection goal the rotation, a deadline under a month keeps the first three.

SurfacePayoffValueLow effortRep gate
Tag excerpt and wiki: routing, links to canonicalscompounding+++++any, peer-reviewed
Self-answered canonical per recurring support themecompounding++++15
New answer on a high-view thread with outdated answerscompounding++++1
Duplicate flags onto a canonicalcompounding++++15 flag, 3,000 vote
New questions answered within 48 h; Staging Ground reviewone-shot++1; reviewer gate on live page
Bounty on a hard, high-view threadone-shot+++75, from the poster's own rep

Answering the backlog oldest-first is deleted, not ranked: the most hours on threads nobody reads. Join a Collective only if your technology already has one.

This order starves reputation-building. Only a gold tag badge (1,000 score, 200 answers in the tag) closes duplicates alone; 3,000 rep is one of three votes, and colleagues never supply the others. Promote badge-building when most of the backlog is repeats.

What the ranking rewards

Published by Stack Overflow:

  • More correct new answers on old threads: accepted answers are no longer pinned above higher-scored ones.
  • Fewer bulk edits: every edit, answer or bounty bumps the thread to the Active tab. Batch weekly.
  • Fewer thin answers: downvoted, closed or deleted posts trigger automatic answer bans.

Practitioner reverse-engineering, unconfirmed: titles quoting the literal error string rank best; high-view threads are what searchers and AI retrievers hit.

Limits

LimitValueType
Flag, upvote, self-answer / comment / create tag15 / 50 / 1,500 repenforced
Edit unreviewed / tag synonym / close vote2,000 / 2,500 / 3,000 repenforced
Links, posting rate, daily flags and voteslimited for new accounts, capped dailyenforced
Six spam or abusive flagspost deleted, -100 repenforced
AI content, undisclosed affiliationdeletion, suspensionenforced on report
Share of own-product answers"some but not all"policed

Thresholds move; check the live privileges page. Reputation is per site; disclosure applies everywhere:

  • Stack Overflow: product answers allowed; "which tool should I use" is off-topic.
  • Software Recommendations: recommendations on-topic if the answer meets every stated requirement.
  • Portuguese, Russian, Spanish, Japanese Stack Overflow: separate tags; answer in that language.
  • Meta Stack Overflow: tag policy and merges after a rename; no promotion.

Cadence

  1. Week 1, route. On your support page and tag excerpt: how-to and error questions to Stack Overflow; bugs, requests, "why" and "when" to your own channels.
  2. Weeks 1-2, new accounts. Answering needs no reputation; start in the product tag. 200 rep on any Stack Exchange site adds 100 on every other.
  3. Week 2, triage the backlog:
    1. Sort by views; ignore the near-zero-view tail.
    2. Cluster by root cause; one canonical each, flag the rest as duplicates.
    3. Deprecated versions: one migration answer as duplicate target.
    4. Bugs posted as questions: workaround, disclosed link to the public issue.
    5. Account or billing: a 50+ rep steward comments the support route and flags to close.
  4. Daily 15 minutes, rotating: tag feed (/feeds/tag/<tag>) into chat; one assignee per question.
  5. Weekly 1-2 hours, steward: Unanswered by views, tag wiki edits, the update batch.
  6. Quarterly: 5-10 canonicals for the top support-ticket themes; search first and answer an existing thread when one exists.

Measure

Targets are self-set; Stack Overflow publishes no vendor-tag benchmark.

  • Answer rate under 30 days (start at 90%); median time to first answer (one business day).
  • Canonical coverage of the top 20 ticket themes; views on canonicals; monthly AI-citation spot check.
  • Hygiene, pass threshold zero: spam flags, AI deletions, undisclosed-affiliation comments, rejected edits.

Output

In order: direct answers to the user's questions; posture with its predicate; ranked actions with owner, hours, rep gate; backlog clusters with canonicals, assumed ones marked; routing table; filled templates with technical bodies left for staff; metric baseline or its query. Read references/templates.md for disclosure lines, canonical Q&A, redirect comments, tag excerpt and wiki, and the Data Explorer query.

References

  • samber/developer-relations-skills@developer-troubleshooting-docs - docs error pages each canonical mirrors.
  • samber/developer-relations-skills@developer-feedback-loop - where bug reports and requests from the tag go.
  • samber/developer-relations-skills@developer-community-health - measuring Stack Overflow beside other channels.
  • samber/developer-platform-skills@github-engagement-playbook - when bugs and requests redirect to GitHub.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.