agentleFS
khadir-syedkhadir-syed/k_ai-agent-skills/skills/socialmedia/post-performance-investigator/SKILL.md

post-performance-investigator

Skill19 starsChanged 52 days agoMITView on GitHub

Investigate why a published social media post performed differently than expected — under or over — using evidence labeled observed/reported/inferred/unknown, and stop before recommending changes. Use when a post's engagement or reach doesn't match expectations; do not use on a post that hasn't been published yet or to draft new posts.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 12 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    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:

    Treat the task as diagnosis-only by default.

    line 12
  2. Distinctive:

    Use only the metrics, comments, and context the user supplies or that is directly available.

    line 13
  3. Distinctive:

    If the user explicitly includes "so what should we post next" in the same request, finish and report the diagnosis first, make the …

    line 17
  4. Distinctive:

    when the post ran and any surrounding context (other posts, promotions, external events).

    line 28
  5. Distinctive:

    Ask only for missing information that blocks a useful investigation.

    line 30
  6. Distinctive:

    Unknown: not established by anything available.

    line 39
  7. Distinctive:

    Do not label the first plausible-sounding factor as the explanation without checking it against the evidence.

    line 54
  8. Distinctive:

    If the evidence cannot distinguish between two or more explanations, report that explicitly rather than picking one.

    line 60
  9. Distinctive:

    Stop after the report. Do not draft a new post or recommend a specific content change as if it were already agreed.

    line 73
  10. Distinctive:

    Stop and explain the boundary when:

    line 77
  11. Distinctive:

    Before returning the report, confirm that:

    line 86
  12. Distinctive:

    no follow-up content was drafted or recommended as a settled decision.

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Post Performance Investigator

Determine the most defensible explanation for why a published social media post performed differently than expected. Produce an evidence-backed diagnosis and stop before recommending any change unless the user separately authorises that.

Operating boundary

  • Treat the task as diagnosis-only by default. Do not draft a follow-up post, edit the original post, or change any scheduling/targeting settings.
  • Use only the metrics, comments, and context the user supplies or that is directly available; do not invent engagement numbers or audience behaviour.
  • Distinguish an actual platform metric from someone's characterisation of it (e.g. "it flopped" is not the same as "reach was 400, versus a 3,000 average").
  • Redact or generalise any real customer names or personal data found in comments before including them in the report.

If the user explicitly includes "so what should we post next" in the same request, finish and report the diagnosis first, make the transition to drafting visible, and follow the host's approval requirements for that separate step.

Investigation workflow

1. Frame the problem

Record:

  • the platform and post in question;
  • the expected performance (a stated goal, a typical baseline, or a comparison post);
  • the actual reported performance and how it was measured; and
  • when the post ran and any surrounding context (other posts, promotions, external events).

Ask only for missing information that blocks a useful investigation. Do not convert someone's impression ("nobody saw it") into an observed fact.

2. Establish observations

For every important claim, label it as:

  • Observed: a specific metric or fact directly supplied (impressions, reach, engagement rate, click-through, posting time, format).
  • Reported: a characterisation given by the user or a stakeholder but not backed by a specific number (e.g. "engagement felt low").
  • Inferred: a conclusion drawn from patterns in other posts or general platform behaviour — not stated by anyone for this post specifically.
  • Unknown: not established by anything available.

State validation limits precisely: a single post's numbers do not prove a trend, and a comment thread's tone does not prove the whole audience's reaction.

3. Compare hypotheses

List the plausible explanations supported by the current evidence — for example, posting time, format mismatch for the platform, algorithm changes, competing content that day, weak hook/caption, or the goal itself being miscalibrated against a realistic baseline. For each hypothesis, record:

  • evidence supporting it;
  • evidence contradicting it;
  • the smallest additional evidence that would distinguish it from alternatives; and
  • its current status: supported, weakened, rejected, or untested.

4. Establish the likely explanation

Do not label the first plausible-sounding factor as the explanation without checking it against the evidence. Separate:

  • Symptom: the performance gap as reported.
  • Likely factor(s): what the evidence best supports.
  • Contributing factors: conditions that plausibly amplified or masked the effect.

If the evidence cannot distinguish between two or more explanations, report that explicitly rather than picking one.

5. Report and stop

Report:

  • investigation status: explained, partially explained, or inconclusive;
  • the leading explanation(s) and confidence;
  • key evidence and its source, labelled observed/reported/inferred/unknown;
  • rejected or unresolved alternatives;
  • validation limits; and
  • a recommended next check, clearly labelled as not yet performed.

Stop after the report. Do not draft a new post or recommend a specific content change as if it were already agreed.

Stopping conditions

Stop and explain the boundary when:

  • the explanation is established well enough to answer the request;
  • available evidence cannot distinguish the remaining hypotheses;
  • the next useful check requires platform analytics access the user hasn't supplied; or
  • the user asks you to also decide what to post next — that is out of scope for this skill.

Quality check

Before returning the report, confirm that:

  • the conclusion explains the performance gap rather than merely restating it;
  • important claims are labelled observed, reported, inferred, or unknown;
  • plausible alternatives were considered or remain explicitly unresolved; and
  • no follow-up content was drafted or recommended as a settled decision.

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.