agentleFS
kayspiegelkayspiegel/survey-insight-skills/.agents/skills/survey-report/SKILL.md

survey-report

Skill2 starsChanged 3 months agoMITView on GitHub

Answers a question about quantitative survey data from a CSV and writes a short, evidence-backed report as Markdown to reports/. Works with any survey CSV (column types are auto-detected). Invoke with your request as the argument, e.g. /survey-report <your question>. Use when the user wants an analysis, numbers, a count, a cross-tab, or a claim checked on survey data.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    All numbers come from the shared engine — never estimate or count "in your head".

    line 22
  2. Very specific:

    Run these from the project root.

    line 29
  3. Very specific:

    Every command accepts --json for machine-readable output.

    line 39
  4. Very specific:

    If a context document sits next to the CSV (<name>.context.md, or survey-context.md when the folder holds exactly one CSV), read it before …

    line 43
  5. Very specific:

    Profile: Run profile to learn n, column indices and types.

    line 54
  6. Very specific:

    If a file already exists, append -2 etc..

    line 69
  7. Distinctive:

    Accuracy over volume: Only claim what the outputs support.

    line 96
  8. Distinctive:

    No personal raw data (e.g.

    line 104
  9. Distinctive:

    Keep it a short report, not a full study.

    line 105

≈ “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

Survey Report

Goal: Answer the question the user asked from a survey CSV with facts, and save a short report as Markdown to reports/.

The user's question is the request they made when invoking this skill. If it contains a file path/name, use that CSV; otherwise use the CSV in the current folder.

Tool

All numbers come from the shared engine — never estimate or count "in your head". Standard library only, no pandas required.

python3 scripts/survey.py <cmd> [...] [--file CSV]

Run these from the project root. If scripts/survey.py is not there, check .claude/scripts/survey.py or locate survey.py in the project.

  • profile — overview: n, every column with index, detected type (META / SINGLE / MULTI / FREETEXT), fill rate. Always run first.
  • freq COL [--by COL2] [--filter "COL=VALUE"] [--top N] — frequency distribution; --by = cross-tab; --filter repeatable (subset). For MULTI columns the percentages sum to > 100% (correct — explain it in the report).
  • text COL [--filter ...] [--sample N] — free-text answers for qualitative themes.

COL = column index (from profile) or an unambiguous part of the column name. Every command accepts --json for machine-readable output.

Survey context

If a context document sits next to the CSV (<name>.context.md, or survey-context.md when the folder holds exactly one CSV), read it before writing. It carries the subject, the glossary, the audience and the study design. Use it for framing and terminology — the figures still come only from the engine. Whatever in it limits the findings, above all the "Who is missing" section, belongs in this report's interpretation and method. Any expectations listed there are something to test, never something to support: never cite one as evidence and never let one shape how a finding is worded. If no such document exists, analyze as before and mention survey-context once at the end of your reply.

Workflow

  1. Profile. Run profile to learn n, column indices and types.
  2. Translate the question into analyses. Pick the right column(s) and run the fitting commands:
    • "how many / what share / distribution" → freq on the target column.
    • "does X differ by Y / relationship / comparison" → freq X --by Y.
    • "… for group Z" → --filter "Z-column=value".
    • "why / what do they value / typical workflow / wishes" → text on a free-text column; with many answers use --sample 40–60, cluster themes, back them with ~2–3 short verbatim quotes. Usually 2–4 analysis commands in addition to profile — a bare distribution is rarely the whole answer. For open questions combine angles (distribution + relevant cross-tab + free-text themes).
  3. Write the report (see format); numbers exactly as the engine returned them.
  4. Save to reports/. Create the folder if needed. File name: YYYY-MM-DD_short-title.md (date via date +%F, title kebab-case from the question). If a file already exists, append -2 etc.; never overwrite.
  5. Tell the user the key takeaway briefly in your reply and give the report path.

Report format (keep it short)

# <question as a concise heading>

**Data basis:** <file> · n = <responses> · report of <date>

## Key takeaway
1–3 sentences answering the question directly.

## Results
- Supported points with **numbers (absolute + %)**. Tables for distributions/cross-tabs.
- For multi-select, note: multiple answers possible, sum > 100%.

## Interpretation
What does it mean? Notable patterns, relationships, caveats (e.g. low fill rate of a
column, small subset after filtering, limits of the study design from the survey context).

## Method
Columns and filters used, in one sentence, so it's reproducible. If a context document was
used, name the recruiting route and who it cannot reach.

Rules

  • Accuracy over volume. Only claim what the outputs support. Percentages relate to the respondents who answered the column (the engine states this).
  • Compute aggregates ("positive overall", top-2-box) from the absolute counts, never by adding the engine's rounded percentages: 73 of 140 is 52.1%, not 13.6% + 38.6%.
  • Statistics in plain text (Chi² = 12.094, p = 0.438, Cramér's V = 0.17), no LaTeX. Pass on the engine's reliability warnings rather than dropping them.
  • Flag small filtered subsets (n < ~30) as weakly reliable.
  • Report language = language of the question (default English).
  • No personal raw data (e.g. individual ids/timestamps).
  • Keep it a short report, not a full study.

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.