agentsclimarketplace
Sign in

Product feedback loop

Skill selamy-labs/agent-skills/skills/product-feedback-loop

Use when turning stakeholder, reviewer, customer, beta, analytics, or interview feedback into product decisions, tracked work, evals, experiments, or an explicit decision to gather more evidence.From its SKILL.md

Install
npx -y skills add selamy-labs/agent-skills --skill product-feedback-loop

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.6 KB, 960 tokens by cl100k_base, as published. Nobody here has run it

Product Feedback Loop

Use this when feedback should change what gets built, measured, or decided. Feedback can come from a stakeholder review, product critique, customer report, beta cohort, support thread, user interview, analytics signal, or sales proxy.

The loop is not "do whatever the latest feedback says." It converts signals into decisions with source lineage, confidence, priority, and verification.

Classify The Signal

For each feedback item, name:

  • Source: reviewer, customer, beta cohort, analytics, support, interview, sales proxy, or other.
  • Evidence level: direct observation, reproduced behavior, measured trend, repeated report, single anecdote, or secondhand proxy.
  • Feedback type: bug, missing capability, UX friction, product strategy, unclear expectation, instrumentation gap, or already-handled item.
  • Affected user path: the workflow, segment, or job-to-be-done involved.
  • Confidence: high, medium, or low, with the reason.

Keep source lineage explicit. Use [[grounded-generation]] when summarizing or combining feedback from external facts.

The Loop

  1. Capture without obeying. Restate the feedback faithfully, but separate the user's words from the product decision.
  2. Cluster by theme. Group repeated reports, related UX friction, and shared failure modes. Do not let one loud anecdote become the whole roadmap.
  3. Map to current truth. Check the current product, docs, issue tracker, PRD, metrics, or code before assuming the feedback is still accurate.
  4. Choose a response. For each cluster, pick one:
    • fix a reproduced bug;
    • improve a confusing flow;
    • add or adjust instrumentation;
    • run an experiment;
    • update the product spec or issue;
    • ask for a decision;
    • gather more evidence; or
    • explicitly defer.
  5. Create the smallest tracked unit. Convert chosen responses into issues, PRD changes, evals, experiments, or implementation slices. Use [[small-focused-changes]] and [[agentic-coding-loop]] for code follow-up.
  6. Add a measurable check. Define the acceptance test, metric, event, rubric, or user-path verification that will show the response worked.
  7. Report decisions and uncertainty. Tell stakeholders what is changing, what is not changing, what needs a decision, and what evidence is missing.

Prioritization

Prioritize feedback higher when it is:

  • frequent across independent sources;
  • severe for a core workflow;
  • blocking activation, retention, revenue, safety, or trust;
  • cheap to verify and fix;
  • aligned with the current product goal; or
  • supported by both qualitative and quantitative evidence.

Prioritize lower when it is:

  • a one-off preference with no supporting pattern;
  • outside the target segment;
  • contradicted by current metrics or source-of-record checks;
  • a feature request that hides an unvalidated product decision; or
  • costly relative to the confidence and impact.

Stop Conditions

Stop with tracked work when:

  • the feedback cluster has a clear response;
  • the response has an owner, artifact, or implementation path;
  • success can be measured or verified; and
  • the decision is consistent with the current product goal.

Stop with a decision request when:

  • two valid product directions conflict;
  • the feedback implies a target-user, pricing, launch, privacy, or safety decision;
  • the confidence is too low for implementation but high enough to investigate; or
  • the requested change would invalidate existing commitments.

Stop with no action when:

  • the feedback is stale, already addressed, out of segment, or unsupported;
  • the right response is to monitor for recurrence; or
  • the cost of action exceeds the expected value.

Output Shape

Feedback cluster: <theme>
Source lineage: <sources and confidence>
Decision: <fix, improve, measure, experiment, ask, gather, defer>
Tracked artifact: <issue, PRD, eval, metric, experiment, PR, or none>
Verification: <how we will know the response worked>
Open question: <none or explicit decision needed>

Anti-Patterns

  • Treating all feedback as equally important.
  • Shipping a change from a single anecdote without naming confidence.
  • Losing the original source while summarizing.
  • Turning product decisions into implementation tickets without owner sign-off.
  • Letting analytics override direct user pain without investigating the gap.
  • Closing the loop with a reply but no tracked artifact or measured outcome.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most product growth skills give in 960 tokens

Counted across 728 of the 1,010 authors here whose files we hold, read 2026-08-07

  • Read product marketing context before asking questionsin 24 of 728, across 18 files
  • Define the ideal customer profilein 21 of 728, across 3 files
  • Document a rollback plan before deploymentin 21 of 728, across 12 files
  • Analyze the codebase to understand the productin 19 of 728, across 1 file
  • Ask clarifying questions about the value propositionin 19 of 728, across 1 file
  • Search for companies matching the criteriain 19 of 728, across 1 file
  • Look for signals of immediate needin 19 of 728, across 1 file
  • Assign a fit score from one to tenin 19 of 728, across 1 file
  • Identify the target decision-maker rolein 19 of 728, across 1 file
  • Suggest a personalized contact strategyin 19 of 728, across 1 file
  • Provide conversation starters for outreachin 19 of 728, across 1 file
  • Format results in a scannable markdown templatein 19 of 728, across 1 file

Said here and by no other author read

  • classify each feedback item
  • keep source lineage explicit
  • restate feedback without obeying it
  • cluster feedback by theme
  • verify feedback against current product records
  • choose one response per feedback cluster

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,452. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.