writing-reports
Skill3 starsChanged 10 days agoMITView on GitHub
Use when the user asks for a report, research write-up, benchmark/model/tool comparison, or any 'give me the report' request.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- 0trm/galloptheir own version, first seen Oct 4, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- 0trm/gallop★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository20 other files it gives its agents.
- AGENTS.mdAGENTS.md
- covidence-full-text-reviewSkill
- covidence-full-text-retrievalSkill
- covidence-screeningSkill
- impeccableSkill
- ilearning-autoadvanceSkill
- oh-my-piSkill
- auditing-worktreesSkill
- camofox-browserSkill
- demo-runbookSkill
- hermes-deepseek-cacheSkill
- i-have-adhdSkill
- memory-uiSkill
- multi-repo-pr-review-merge-roundSkill
- oh-my-ghosttySkill
- tldrSkill
- sci-hub-accessSkill
- slack-mention-watchSkill
- slack-scanSkill
- superpowers-using-superpowersSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Not for: one-line answers, PR descriptions (pr skills), Covidence verdicts (own skill).
- Distinctive: line 29
Compute: All derived metrics (blends, ratios, per-point costs) calculated in Python, never mentally.
- Very specific: line 50
When the caller supplies the data and forbids web search, the ≥2-source cross-check is waived — but the Method line must say Inputs as …
- Very specific: line 54
Every report file lands in ~/Projects/auto-learn-for-me/reports/YYYY-MM/ as YYYY-MM-DD-<slug>.md (e.g.
- Distinctive: line 56
When a report is requested in chat, still write the file (chat is ephemeral.
- Distinctive: line 57
If a report was written elsewhere (a work repo, another machine path), move or symlink it into the library and note the original location …
≈ “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.
Files
Writing Reports
Overview
A report is a decision document: verdict first, evidence second, provenance always. Every report ships in two parts: a dated markdown file (full) + a short chat lead (verdict + key table + file path).
When to Use
- "give me a report" / research-then-writeup requests
- Benchmark, model, tool, price, or data comparisons
Not for: one-line answers, PR descriptions (pr skills), Covidence verdicts (own skill).
Method
- Gather. SearXNG search first (if :8888 down → restart-first rule, skill selfhosted-search). Scrape sources with curl + parse programmatically. Never single-source a number — cross-check prices/benchmarks in ≥2 independent sources (waived only in Trusted-Input Mode). Date-stamp: fetched data gets its fetch date, computed-only data gets the compute date.
- Compute. All derived metrics (blends, ratios, per-point costs) calculated in Python, never mentally. State the formula next to the column (e.g. "blend = 3:1 in:out").
- Write the file.
~/Projects/auto-learn-for-me/reports/YYYY-MM/YYYY-MM-DD-<slug>.md(see Report Filing) — skeleton inreferences/report-template.md. - Deliver. Chat = verdict + the one key table + file path. If the user wants it in Word → plain text, ALL-CAPS headers, no markdown, piped to
pbcopy.
Report Structure
# Title → metadata bullet list (Date / Scope / Method — one - **Label:** value bullet each) → Verdict (first sentence of prose) → tables → numbered findings → caveats → numbered references. Inline [n] citations throughout. Fill the template; don't improvise structure.
Hard Rules
- Verdict in the first sentence of prose. The metadata block comes first in the file (template order); the Verdict section opens with the decision — no "Executive summary" preamble, no restating the request.
- Every number has a source or a formula. Measured vs vendor-claimed vs derived — say which. Caller-supplied figures are labeled
caller-supplied(not vendor-claimed). - Assumptions are labeled
ASSUMPTION:inline where used, never hidden in a footer. - Cite inline, number the references. Every sourced claim, number or quote carries an inline marker right after it:
[n], or[n, m]for several sources. The## Referencessection lists each source once, numbered in order of first citation:[n] Title: <URL>. That is the title (with publisher and date when useful), then the URL in angle brackets. Don't use bare URLs, and don't invent titles for unfetched URLs (use[n] [unfetched — <URL>]). The docs site turns each marker into a linked superscript with a hover preview of the title, and links each reference back to its citations. Uncited references and citations with no matching reference are both errors: check that the set of cited numbers equals the set of reference numbers before filing. - No
---separators between sections. Headings separate sections. Rules drawn between every section make a report harder to read. - No invented scenarios. Illustrative math (e.g. monthly cost) must be flagged as an estimate with its basis.
- Chat lead follows ADHD shape: answer first, tables over prose, no preamble/recap (skill i-have-adhd).
- Filing is the default. Skip the file only when the caller explicitly forbids writing files; then say in the reply that the report is chat-only and unfiled.
Trusted-Input Mode
When the caller supplies the data and forbids web search, the ≥2-source cross-check is waived — but the Method line must say Inputs as supplied by caller, not independently verified, and Caveats carries the single-source risk. Never silently skip verification.
Report Filing (dated reports library)
Every report file lands in ~/Projects/auto-learn-for-me/reports/YYYY-MM/ as YYYY-MM-DD-<slug>.md (e.g. 2026-09-08-commandcode-cost-efficiency.md) — this folder is the running library of past reports, so a later "that report we did on X" resolves to a file, not session memory.
- When a report is requested in chat, still write the file (chat is ephemeral; the file is the artifact).
- If a report was written elsewhere (a work repo, another machine path), move or symlink it into the library and note the original location in its metadata block.
- Need something from an old report?
ls reports/YYYY-MM/or search by slug before re-researching.
Common Mistakes
| Mistake | Fix |
| "Executive summary" heading before the verdict | Verdict IS the first line |
| Bare URL in references | Numbered [n] Title: <URL> per source |
| Sources listed but never cited in the text | Inline [n] after each sourced claim; cited set = reference set |
| Metric column with no formula | Formula beside the column header or footnote |
| Unlabeled invented usage scenario | ASSUMPTION: inline + basis |
| Full report dumped into chat | File + short lead; chat table ≤ 10 rows |
| Report only in chat, no file | Chat is ephemeral — file it: reports/YYYY-MM/YYYY-MM-DD-<slug>.md |
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
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.

