agentleFS
hsigstadhsigstad/research-kit/skills/gmail/SKILL.md

gmail

Skill0 starsChanged 4 months agoMITView on GitHub

Draft an email for Henrik in his house style AND record the draft so the sent-vs-draft diff can teach the style. Use when any session drafts/prepares an email for Henrik to send (to a coauthor, colleague, admin, etc.), or when the user says 'draft an email', 'reply to <person>'.

28 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use google_sheets, google-chat and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Shugar86Skill

    google_sheets

    Google Sheets read/write integration

    Also in 3 projects that use this

  • antoniolgSkill

    google-chat

    Read Google Chat spaces/threads via the Chat API, create/refresh OAuth tokens, parse Gmail Chat URLs, and list spaces. Use when asked to look up or verify messages in Google Chat or when you need API access to Chat from the CLI.

    Also in 3 projects that use this

  • kennyolofsson23-netizenSkill

    elevenlabs

    ElevenLabs voice AI integration. Text-to-speech, speech-to-text, voice cloning, and audio processing via ElevenLabs API and MCP.

    Also in 3 projects that use this

  • ansvisorSkill

    shopping

    Checks product pages and feed for ChatGPT Shopping (US) and Google Shopping (EU).

    Also in 3 projects that use this

  • ourinesSkill

    forget

    Delete saved knowledge from ~/.claude/learnings/ and optionally clean up knowledge graph.

    Also in 3 projects that use this

  • G-HunterAiSkill

    tiktok

    TikTok platform integration. Manage videos, view analytics, and track engagement.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 24 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read research/meta/email_style.md first and draft to it (short · link-don't-inline · Gmail-default · no em dashes · terse concrete …

    line 15
  2. Very specific:

    Write the draft to research/meta/email_drafts/YYYY-MM-DD_<session>_<slug>.md — one file per draft, frontmatter + body:

    line 20
  3. Very specific:

    Keep drafts out of any *public* repo body if they contain private personal info.

    line 58

“Common” is measured: kept by at least 6 in 10 of the 24 edited versions of this file we could read. “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

/gmail — draft an email + record it for the style-learning loop

Any session that prepares an email for Henrik uses this, so (a) the draft follows his house style and (b) it is recorded at draft time — which lets Saga diff Henrik's sent version against the draft and learn his edits, no matter which session drafted it. (The outlook skill only archives already sent .eml threads; this records the draft before it's sent.)

1. Draft to the house style

Read research/meta/email_style.md first and draft to it (short · link-don't-inline · Gmail-default · no em dashes · terse concrete scheduling · don't over-explain an accept · …). Do not invent a different style; that file is canonical and is what the learning loop tunes.

2. Record the draft (the important part)

Write the draft to research/meta/email_drafts/YYYY-MM-DD_<session>_<slug>.md — one file per draft, frontmatter + body:

---
date: 2026-08-13
session: <your session name>        # e.g. Saga, deterrence, judgeGPT — the attribution + context key
to: <recipient email>
subject: <subject>
thread_id: <gmail thread id if replying, else "">
project: <slug or "">               # e.g. deterrence, vague, "" for admin/general
why: <one line of context NOT obvious from the email — what prompted it / what it's really about>
sent: false                          # Saga flips this after diffing against Henrik's sent mail
---

<the exact draft body>

The why line is also the cross-session context record — so a later reader (usually Saga during reconcile) isn't guessing why the email exists or who drafted it.

3. Create the actual draft — only if you can

  • If you have the Gmail MCP (Saga does): create the Gmail draft (create_draft, replyToMessageId when replying; htmlBody <a> anchor for any link per email_style rule 10), and add its draft id to the recorded file (draft_id: r-…). Leave a Saga/handled label if that's your convention.
  • If you don't (most research/teaching sessions): just record the file (step 2) and hand Henrik the body to send himself. Never send — the outward act is always Henrik's.

4. The learning loop (Saga runs this — you don't)

Saga's nightly/triage scans meta/email_drafts/ for records whose thread now has a SENT message from Henrik, diffs sent-vs-draft, appends the delta to meta/email_feedback_log.md, sets sent: true, and periodically distills recurring deltas into meta/email_style.md. So every session's draft feeds the same house-style improvement — the whole point of recording here rather than in a per-session file.

Notes

  • Sensitive routing (site-password allowlists, delivery quirks like a UiO spam-reject, WhatsApp-only contacts) is Saga's overlay in its private repo — not in this shared file. Draft addresses come from contacts.yaml default_email.
  • Keep drafts out of any public repo body if they contain private personal info; research/ (this meta-repo) is private, which is why the archive lives here and not in a project's shared repo.

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.