Systems Lab

Agent skill

capture-feedback

Turn something heard in the field into durable context, at the right confidence level.

activeSelf-containedInstructions only352 words

Filed under Prospecting and list building.

From getcargohq/cargo-manifest · 4 skills · 6 · pushed 2026-09-12

What it does when it runs

Turn something heard in the field into durable context, at the right confidence level. Use when the user relays an objection, a competitor mention, a win or loss reason, a signal, or any claim from a call, and asks to remember it. Enforces the repetition rule and the confidence ladder.

Read from the skill and the 0 files bundled beside it. A skill’s own description is written to be selected by an agent, so it describes the job and not the dependencies.

Keys and connectors you must supply
None found.
Hosts it reaches
No third-party host appears in the skill or its bundled files.
Tool permissions it declares
No allowed-tools in the frontmatter. It only issues instructions, so there is nothing to bound.
Actions present in the files
None. Instructions only.

Ask about capture-feedback

Opens your assistant with this page's verified links already in the prompt.

Is this safe to install?ClaudeChatGPT
Adapt it to my stackClaudeChatGPT
What else do I need for it to workClaudeChatGPT
Rather ask a human? Talk to Cheetah

Install it

View source on GitHub ↗
git clone --depth 1 --filter=blob:none --sparse https://github.com/getcargohq/cargo-manifest.git /tmp/cargo-manifest
git -C /tmp/cargo-manifest sparse-checkout set ".claude/skills/capture-feedback"
mkdir -p ~/.claude/skills/capture-feedback
cp -R "/tmp/cargo-manifest/.claude/skills/capture-feedback/." ~/.claude/skills/capture-feedback/

Picked up without a restart. A project skill of the same name is shadowed by your personal one. For one repository only, swap ~/.claude/skills for .claude/skills. Claude Code docs ↗

The folder is the same in every client that implements the format — 46 of them — so if yours is not above, only the destination changes.

The skill

Source on GitHub ↗

Reproduced in full from getcargohq/cargo-manifest/blob/c8b9850b5f9cf07d5a2290dd42227e2ac128a906/.claude/skills/capture-feedback/SKILL.md, which is licensed MIT (repository). 352 words, 4 headings.

Capture feedback into context

The context layer is only worth what its weakest claim is worth. This skill is the gate.

The two rules this enforces

  1. Repetition. A claim enters context/ after it shows up twice, independently. Once is an anecdote and belongs in cadence/log/ or context/insight/ at confidence: hypothesis.
  2. The confidence ladder. Every file in proof/, signal/, and insight/ carries confidence::
    • hypothesis : we think so. Readable, but outreach plays may not cite it. The linter blocks that.
    • validated : seen at least twice, independently, in the real world.
    • proven : measured and repeatable, with the number to show.

Steps

  1. Classify what you heard. Which domain does it belong in?

    • A pushback in a sales conversation → objection/
    • A number, quote, or customer result → proof/
    • A buying trigger → signal/
    • A market observation that fits nowhere else → insight/
    • A competitor or status-quo alternative → alternative/
    • A won or lost deal and why → client/

  2. Search for prior occurrences. Grep context/, recent cadence/log/ entries, and outputs/. Has this shown up before?

  3. Decide the confidence level, honestly.

    • Second independent occurrence, no numbers → create or upgrade to validated, and cite both occurrences in the body.
    • You have a measured, repeatable number → proven.
    • First occurrence → either log it (preferred) or write it at hypothesis. Say out loud that it is not yet citable by outreach.

  4. Write it. New file from the domain's _template.md, or upgrade the existing file's confidence: and add the new evidence. Cross-reference related files with [[domain/slug]] wikilinks so the knowledge graph connects.

  5. Never overwrite a major claim silently. If you are rewriting an existing position, archive the prior version rather than deleting it.

  6. Lint. Run npm run lint:context. Then report what you wrote and at what confidence.

Rules

  • Do not invent evidence to justify an upgrade. An honest hypothesis beats a fabricated validated.
  • Upgrading a claim to validated unlocks it for outreach copy. Treat that as the consequential act it is.

Other skills for the same job

Different authors, same problem. Matched on the words in the skill name, across every library in the catalogue except this one.

Need help setting it up?

This page tells you what capture-feedback does and what it needs. Cheetah builds the agent setup it runs inside: data, CRM, sequencing and the guardrails.

Book a call →

The directory stays free. There is nothing gated behind this.