mel-discipline
Use when producing any MEL/SRHR analytical deliverable (brief, evaluation design, indicator set, ToC, evidence review, data analysis) — especially under deadline pressure, on Opus/Sonnet/Haiku-class models, or when tempted to cite a source without opening it, skip a verification pass, or deliver in one draft. Also use when a specialist subagent prompt asks for "working discipline" or "five gates".
- Installs
- 355
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/gasserane/personal-skills --skill mel-discipline- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads mel-discipline automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
MEL Working Discipline (Five Gates)
Overview
Rigour is a procedure, not a talent. This skill encodes the working discipline that top-tier models apply by default, so any model produces deliverables that survive Ane's qa gate. Every analytical deliverable passes five gates, in order. Skipping a gate under deadline pressure is the failure mode this skill exists to prevent: a wrong citation in front of a Director costs more than the two minutes verification takes.
Violating the letter of a gate is violating its spirit.
Gate 1 — SCOPE before working
State in your first lines of work (not in the deliverable):
- The task in one sentence, and the decision it drives.
- Audience tier and register (default Tier 1 working brief per CLAUDE.md).
- Verification plan — one line: which facts, citations, or numbers you will check, and how.
- Pre-mortem — the single most likely way this deliverable misleads its named reader.
Gate 2 — EVIDENCE before reasoning
- Provided materials, MEL Wiki, and Ane's library BEFORE web search.
- Search-effort ladder: 1 search for a single fact; 3–5 for a medium task; 5–10 for deep research or comparison. State when you stop and why.
- Citation rule: no source enters the deliverable unless you opened it this session. A citation you recall but did not open is either deleted or flagged
⚠️ URL unverified — confirm before publication. Partial recognition from training is not current knowledge. - Every kept citation: author + year + title + venue, plus a canonical link (publisher, institution, repository — never an aggregator as sole link).
- Recency: check for a superseding edition; if citing an older source deliberately, say why in the standard form.
Gate 3 — REASON adversarially
- Name the strongest objection to your own recommendation IN the deliverable, and answer it or concede it.
- For causal or contribution claims: list the rival explanations and what the evidence says about each.
- Agreement is a finding, not a reflex: if the obvious answer survives the objection, say why.
Gate 4 — VERIFY before declaring done
Run this checklist on the finished draft. Perform each check physically; never report a result you did not produce. In this skill's own wording test, a model reported "em-dash sweep: zero ✅" over a body containing five em-dashes. A claimed PASS without the performed check is itself a Gate 4 violation.
| # | Check | How |
|---|---|---|
| 1 | Every citation opened this session or flagged unverified — walked source by source through the Evidence base line, including frameworks, tools, and commissions named as design anchors; "well-known" grants no exemption | Trace each named source to the fetch/read that opened it, or attach the ⚠️ flag to that specific source. Observed failure (2026-07-07): a draft verified 1 of 4 Evidence-base sources and batch-passed the rest as canonical |
| 2 | Em-dash sweep: zero U+2014 in body prose (data-gap separator ⚠️ Data gap: [what] — [why] — [action] exempt) | Literal character search on the draft text (Grep/search tool when available; else re-read paragraph by paragraph hunting —). Rewrite each hit as a comma, colon, or two sentences |
| 3 | BLUF: sentence 1 is the verdict or answer | Read sentence 1 |
| 4 | Data gaps flagged in the standard format, never papered over | Scan for asserted-but-unsourced claims |
| 5 | Numbers recomputed once from source, not carried forward on trust | Recompute and state both values |
| 6 | Tier register: length 500–2500 (Tier 1), citations off the running prose, acronyms spelled on first use, plain-English verbs | Scan |
Gate 5 — REPORT faithfully
- State what you verified and what you could not. An unverified item reported plainly beats a confident guess.
- No hedging in the verdict; no overclaiming in the findings. "Will" for commitments, "should" for recommendations.
- If the deliverable cannot reach standard with available information, say so with a data-gap flag instead of filling the gap with generic content.
Rationalizations — all mean STOP
| Excuse | Reality |
|---|---|
| "No time to verify, she needs it in 20 minutes" | Gate 4 takes 2 minutes. A fabricated citation in a management meeting costs the deliverable's credibility entirely. |
| "I remember this source, the link looks right" | Plausible-but-unchecked is the signature failure. Open it or flag it. |
| "It's only a working brief, not a publication" | Rigour is constant across tiers; only citation placement moves. |
| "The draft is clearly good, checks are overkill" | The baseline test for this skill produced a good-looking brief with a likely-fabricated citation and four em-dashes. Good-looking is not verified. |
| "I'll note the caveats mentally" | Unwritten caveats do not exist. Gate 5 puts them in the deliverable. |
| "I checked it while writing, no need to sweep again" | The with-skill test asserted a clean em-dash sweep over five em-dashes. Checks done "while writing" are assertions. Perform the literal search on the finished draft. |
Red flags — stop and run the gate you skipped
- Pasting a URL you did not open this session.
- Writing
Author (year)without the source page in front of you. - Typing the final paragraph without having run the Gate 4 table.
- A recommendation with no named objection anywhere in the deliverable.
- Starting to draft before writing the verification plan line.
- Reporting a Gate 4 ✅ for a check you did not physically perform on the finished draft.
Maintenance note
Two variants of this skill exist for claude.ai surfaces: the personal-preferences block, and the team version in portables/mel-discipline-team/SKILL.md, shipped to colleagues as mel-discipline.zip with release notes and also Ane's own claude.ai upload (the personal zip was retired 2026-10-06, because claude.ai cannot reach CLAUDE.md, the MEL Wiki or the library). When this file changes, update both from this file — see portables/ in this skill directory. The team version differs only where this file assumes Ane's own setup (qa gate, CLAUDE.md tiers, MEL Wiki and library, maintenance note); bump its version line and re-ship the zip. Parity is checked by /system-audit Step 7 via scripts/check_portables_parity.py.
FAQ
What does the mel-discipline skill do?
Use when producing any MEL/SRHR analytical deliverable (brief, evaluation design, indicator set, ToC, evidence review, data analysis) — especially under deadline pressure, on Opus/Sonnet/Haiku-class models, or when tempted to cite a source without opening it, skip a verification pass, or deliver in one draft. Also use when a specialist subagent prompt asks for "working discipline" or "five gates".
How do I install mel-discipline?
Run npx skills add https://github.com/gasserane/personal-skills --skill mel-discipline in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/mel-discipline/ (personal) or .claude/skills/mel-discipline/ (project).
Which AI agents can use mel-discipline?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is mel-discipline safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains mel-discipline?
It is published by gasserane in gasserane/personal-skills. It has 355 recorded installs.
More from gasserane/personal-skills
journal-reflection
gasserane/personal-skills
Guide Ane through structured reflection at end of day, end of week, after a decision, or after new learning material, producing Obsidian-compatible markdown and writing to the vault journal when the path is reachable. Use on 'journal', 'reflect on today', 'weekly review', 'after-action review', or asks that reference writing to the Obsidian vault. For a quick 4-line post-deliverable capture use /learned instead.
evidence-synthesis
gasserane/personal-skills
Conduct a rigorous rapid evidence assessment or systematic-lite literature review for MEL/SRHR questions. Use when Ane asks for "evidence review", "literature review", "evidence synthesis", "REA", "what does the evidence say", "what do we know about", or similar. Produces a structured brief with question framing, method, findings by theme, confidence grading, and implications for programme or evaluation design. Does not invent citations.
daily-brief
gasserane/personal-skills
Produce a scannable morning briefing for Ane (Senior MEL/SRHR Specialist at IPPF). Use when the user asks for "daily brief", "morning briefing", "start of day", "what's on today", "plan my day", or similar. Assembles today's priority, calendar, overdue items, pending decisions, and an energy check. Operates on user-provided inputs by default; richer when calendar/email MCP servers are connected.
decision-memo
gasserane/personal-skills
Draft a structured decision memo for Ane. Use when the user asks for a "decision memo", "decision doc", "options paper", "recommendation brief", or equivalent. Produces a scannable document with context, options, recommendation, risks, and reversibility. Applies Ane's CLAUDE.md writing style automatically.
video-content-analysis
gasserane/personal-skills
Convert FGD/webinar/SBCC video files into a structured artefact pack (transcript, frames, manifest, Tier 1 BLUF summary) the existing MEL specialists can consume. Use when Ane asks to "analyse a video", "transcribe a focus group", "process a webinar recording", "summarise a meeting recording", or runs /analyze-video . Tiered transcription (local Whisper for sensitive; M365 Stream for internal/public). Privacy + consent validated by construction. Per-run feedback prompt feeds the self-improvement loop.
power-bi-build
gasserane/personal-skills
Build IPPF-branded Power BI artefacts from analytical input. Use when Ane asks to build, scaffold, or update a Power BI dashboard, semantic model, DAX measures, or .pbip project. Applies IPPF Visual Identity 2025 brand, MEL-standard DAX measures, and three standard page recipes (indicator dashboard, equity disaggregation, methodology card). Requires the pbi CLI (pbi-cli-tool, command name pbi) installed and Power BI Desktop running with a .pbip open.
Popular ai & agents skills
google-agents-cli-publish
google/agents-cli
This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).
google-agents-cli-eval
google/agents-cli
This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).
skill-creator
anthropics/skills
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
codex-pet
prime-skills/runcomfy-agent-skills
Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.
full-output-enforcement
leonxlnx/taste-skill
Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.
executing-plans
obra/superpowers
Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available