Personal Skills

Li — Knowledge Manager for Ane's library and MEL Wiki. Use when Ane needs to catalog, retrieve, or reorganize documents in the personal knowledge library, or query/maintain the MEL Wiki. Handles INGEST, QUERY, LINT, PROGRAMMES, and Obsidian vault-survey operations. Does not answer domain questions — retrieves and organizes knowledge for other agents and Ane.

Vi — HR Specialist and Execution Orchestrator for MEL/SRHR work. Receives an approved plan from Ann (or directly from Ane), designs the specialist roster, spawns specialists as subagents, reviews their outputs, compiles the final product, and returns it. General-purpose — invoked by Ann via Agent tool, or directly by Ane when a plan is already approved.

Ann — Convener for MEL/SRHR work. Use when Ane brings any analytical, evaluation, SRHR, or structured-output task. Ann classifies task complexity, queries the MEL Wiki, retrieves knowledge, creates an implementation plan (verifies with user for complex tasks), coordinates with Vi for execution, runs a 5-point quality gate, and delivers. General-purpose — not tied to any specific project.

Fast post-deliverable learning capture. Use when Ane types "/learned", or says "capture what I learned", "log a learning", "what did I learn from this", or wants a quick 4-line note after finishing a piece of work. Appends a dated entry to the running learning log in the Obsidian vault. The explicit-command sibling of journal-reflection's Post-deliverable capture mode; for end-of-day, weekly, decision-review, or new-framework reflection, use journal-reflection instead.

Session wrap-up. Use when Ane types /wrap-up to close a session. Gathers git, errors, and pending actions, delivers a concise status report, then commits and pushes pending work, and offers opt-in post-commit follow-ups (skill improvements, skill-fit signals, learning capture). Pass "report" for a read-only status with no commit. Writes a continuation handoff prompt for the next session whenever named work is still open, one file per workstream so parallel sessions cannot overwrite each other; pass "continue" to force one.

Produce an MA-facing deliverable in Romanian, French, Spanish, or Russian to IPPF rigour, with English as the canonical source: curated SRHR glossary, register preservation, safeguarding-aware term choices, back-translation QA, native-review flags. Use for any localise, translate, or review-my-translation ask, including slide-deck or Storyline exports to Word where the translatable text sits in slide images. Never authors from scratch in the target language. Not for instrument design (instrument-designer), English voice audits (ane-voice), or research video analysis (video-content-analysis).

Generate a donor proposal pack (CERV first) with the full MEL stack: intake, a fixed roster delegated to Vi (evidence-synthesis, toc-builder, indicator-designer, the gender and safeguarding lens specialists, proposal-architect), then the seven-artefact branded pack with the AI-disclosure colophon. Use when Ane builds a grant proposal or runs '/proposal --donor cerv'; donor is an argument so Gates/OSF/UN are later --donor values. Does not fill Part A portal forms, submit, or sign off finance/legal. Distinct from donor-proposal-scoring (scores a written proposal) and implementation-pack (post-award).

Audit and rewrite AI-generated or AI-edited prose to match Ane's IPPF MEL/SRHR voice, preserving rigour: strips hedging, filler, nominalisations, em-dashes, passive voice, abstract openings, misplaced framework name-dropping, and inline citations where an evidence-base line belongs. Tier-aware, defaulting to the Tier 1 working-brief register. Use when the user pastes text and asks to humanize, de-AI, sharpen, tighten, fix the voice, or review the draft's tone. Does not push prose toward casual or blog tone.

A relentless one-question-at-a-time interview to stress-test a MEL/SRHR design before building it: a theory of change, evaluation design, indicator set, results framework, proposal angle, or learning question. Walks the MEL branches (assumptions tested vs untested, OECD-DAC criteria, output/outcome/impact, disaggregation, lens application, data gaps, audience tier), recommends an answer for each question, and flags any fact it cannot source. Use when Ane says "grill my ToC", "stress-test this evaluation/indicator set/proposal", "pressure-test my design", "interview me about this", or wants to sharpen a MEL design before drafting. This is the MEL-lane grill for any MEL or SRHR design. Not for a full Ann orchestration (/ann), a code build (brainstorming), generic non-MEL plan sharpening (/grill-me or /grilling, both user-invoked only), or a contested call Ane cannot answer herself (/decision-debate).

Search 250M+ scholarly works via the OpenAlex API directly: quick search, duplication test, debate map (anchor papers), forensic sourcing of one specific claim, era split across a year, plus a Consensus escalation for design-filtered searches (study type, sample size, journal tier). Use when Ane wants published literature found, a claim sourced, anchor papers, or to know whether a question has been done already. For a full structured evidence review use /evidence-synthesis; for a COMPLEX-tier Evidence Brief use /researcher. Not for reference management (Zotero via /li) or Ane's own library (mcp knowledge search).

Researcher — Evidence Synthesis Specialist. Use when a complex MEL/SRHR task requires deep evidence synthesis before planning begins. Triggered by Ann between PHASE 1 and PHASE 2 for COMPLEX tasks. For a standalone rapid evidence assessment or lighter literature review, use /evidence-synthesis; reserve this skill for the deep COMPLEX-tier Evidence Brief Ane invokes directly when she needs that depth.

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.

Build a rigorous Theory of Change for MEL/SRHR programmes following Vogel (2012) DFID criteria and van Eerdewijk et al. (2017) KIT feminist ToC model. Use when Ane asks for a "theory of change", "ToC", "programme logic", "change pathway", "pathway analysis", or similar. Produces articulated outcomes, preconditions, assumptions (tested vs untested), contribution questions, and an evidence plan. Flags missing feminist political economy analysis.

Apply substantive quality and reliability grading to Ane's MEL/SRHR system (Ann/Vi/Li/Researcher team + wiki + harness + three-repo architecture): letter grades on 9 components and 10 dimensions plus OECD-DAC adapted criteria, three concrete risks, what an external evaluator would still want, and recommendations for moving the grade, with a mandatory self-grading bias note. Use when Ane types /grade-system, /grade, /quality-grade, /rate-system, or asks how good the system is. Honest grading, not flattery. Different from /system-audit (finds specific issues to fix; this grades overall quality posture).

Run a quality, sanity and efficiency audit of Ane's MEL/SRHR system (Ann/Vi/Li/Researcher team, wiki, harness, three-repo architecture): surfaces drift, inconsistencies, hygiene issues and architectural ceilings, categorised by severity, as a Tier 1 working brief with sequencing. Does NOT auto-execute fixes. Use when Ane types /system-audit, /audit, /system-check, /system-health, or asks whether anything is broken or drifting; run before any /li curate and after agent-improvement work. Different from /test (harness only), /li lint (wiki only), and /grade-system (overall quality posture rather than specific issues).

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.

Turn Ane's meetings, 1-1s and document review cycles into structured IPPF-branded meeting notes from any capture method (dictation, notebook photo or scan, recording, Teams transcript, margin comments/marginalia in a .docx), and prepare the next one. Use for meeting notes, minutes, 1-1 debrief or prep, or extracting decisions from a commented document. Not for importing typed claude.ai sessions (capture-desktop), options papers (decision-memo), personal reflection (journal-reflection), daily planning (daily-brief), research video analysis (video-content-analysis), or editing the reviewed document itself (office-review-pass; tor-procurement finalise for a ToR).

Handle a hand-edited Office artefact that is now the authoritative version: repair a feature-rich Excel workbook in place through COM (broken slicers or dropdowns, #REF!, circular reference warnings), or promote a generated .docx/.pptx/.xlsx to canonical hand-edited source and retire its generator. Use when Ane reports a damaged workbook or declares an edited file the source of truth. Distinct from office-review-pass (Word review cycle), the ane_package.reporting builders (generate new branded files), and check-deliverable (read-only QA verdict on prose).

Use when Ane types /improve-system, asks to "mine recent sessions", "what is the system getting wrong", "improvement loop", "trend the rejection rate" — or asks "where are my tokens going", "token audit", "why was that session expensive" (spend-audit mode) — or when the SessionStart maintenance nudge reports improve-system due (14-day cadence). Mines the QA logs and session history for recurring correction patterns, proposes system edits, and audits where token spend goes. Not for one-off bug fixes (/system-audit finds those), not for wiki integrity (/li lint), not for quality-posture grading (/grade-system).

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".

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.

Capture a claude.ai Desktop session into the MEL file system. Use when Ane finishes a session in the Claude Desktop app and wants the work captured on the device — triggered by "/capture-desktop", "capture this desktop session", "I did this on Claude desktop", "save my desktop work", "import from claude.ai". Paste-driven; no cloud API exists. Stages a local transcript, classifies content into deliverables, decisions/rules, project-knowledge changes, and full log, then files each to its home. Auto-files transcript + deliverables; confirms memory/wiki/CLAUDE.md writes one-by-one.

Run a formal multi-perspective debate on one contested MEL/SRHR design decision: persona agents mapped to real specialist lenses research the call independently and blind, debate one round, then a decision-memo-shaped synthesis keeps agreements and genuine disagreements visible for Ane to decide on. Use when Ane cannot decide between named options and wants independent perspectives argued. Token-heavy by design: reserve for real decisions. Not for sharpening a design she already leans on (/grill-mel), net-new builds (brainstorming), or a plain options paper (/decision-memo), and never for building anything.

Produce one of three learning-agenda output types: canonical federation agenda (v1), 1-page quarterly RAG dashboard (v2), or MA adaptation template (v3), operationalising the IPPF EN A+C hybrid locked 2026-05-10. Reuses verified citations from the wiki only; never fabricates sources; applies the IPPF Visual Identity 2025 brand. Use when Ane asks for a learning agenda, annual learning questions, the RAG dashboard, or the MA adaptation template.

Build, review, grade and defend procurement Terms of Reference (ToRs) for consultancy and service contracts: build a branded docx, grade 1-10 with strong and weak points, respond to a named reviewer's objections, or finalise revisions with edit-preservation. Use when Ane drafts, hardens, or defends a ToR, or prepares to contract a consultant or service provider. Distinct from /proposal (funding proposals TO donors), implementation-pack (post-award tracking), procurement-offer-review (incoming supplier offers), and accreditation-desk-review (MA compliance). When Ane is the reviewer commenting on someone else's draft, that is the opposite direction and not this skill.

Answer questions about evaluating GenAI / AI tools and products using the AI Evaluation Playbook (eval.playbook.org.ai; The Agency Fund with CGD and IDinsight, 2026) — the 4-level framework (Model / Product / User / Impact), Minimum Viable Evaluations, golden datasets, LLM-as-judge, engagement funnels, guardrail metrics, level linkages. Use when Ane asks how to evaluate an AI tool, chatbot or GenAI product, "how do I know this AI is working", "what should we measure for this AI tool", "where do I start with AI evals", or an MA/partner AI tool proposal needs an evaluation appraisal. NOT for programme or MEL evaluation design (/ann, /grill-mel, indicator-designer, evidence-synthesis own that lane) and NOT for AI used by evaluators as a working tool (wiki page ai-mel-framework governs that).

Convert a Researcher Evidence Brief (one slug under literature-reviews/) into one of four learning products: MA-facing 2-page brief, workshop / CoP pack, regional learning brief, or donor / management talking-points memo. Reuses the brief's verified sources only; never fabricates citations; applies IPPF Visual Identity 2025 brand and the Tier 1 working-brief register. Use when Ane asks for a learning product or names one of the four output types from an Evidence Brief.

Turn an external tech/AI source (YouTube video, podcast, blog article, tweet or thread) into a de-duplicated improvement roadmap for the MEL system and Claude Code setup: fetch the content, extract concrete techniques, inventory what the system already does, classify each idea reuse/improve/rebuild on an Adopt/Trial/Assess/Hold ring, and write the roadmap file with a mandatory What-NOT-to-build section. Use for "analyse this video/article/talk/thread for my system", "turn this into a system-improvement plan", or a pasted URL with improvement intent. Not for MEL research video files needing the transcription+consent pipeline (video-content-analysis), mining internal QA logs (/improve-system), inspecting the system itself (/system-audit, /grade-system), academic evidence review (evidence-synthesis), or writing the LinkedIn piece (linkedin-field-note).

A fast single-pass scoping map of a MEL/SRHR question: which perspectives it demands, where those perspectives conflict, and whether it is worth a full orchestration, ending in a triage verdict, never citations or findings. Use when Ane wants a question, proposal angle, learning question, or evaluation framing scoped or triaged before committing tokens to a full run. Not an interview of Ane (/grill-mel), not the deliverable or any sourced evidence (/ann, /researcher, /evidence-synthesis), and not a multi-agent debate (/decision-debate).

Run the qa-reviewer maker-checker gate against one finished text deliverable without a full Ann orchestration: spawns qa-reviewer alone on a single markdown or text file, sets the audience tier (default Tier 1 / colleague; override with --tier and --subgroup), and returns a BLUF verdict plus the full qa_block. Read-only: rewrites nothing, writes nothing to disk. Use on '/check-deliverable <file>' or any ask to QA a finished brief on its own. Refuses Office binaries (.docx/.xlsx/.pptx). Not for producing a deliverable (/ann), voice rewriting (/ane-voice), or code review (/code-review).

Build the Excel toolkit that runs a selection between a published ToR and the award: procurement mode (master workbook plus one scorer workbook per panel member) and appraisal mode (generic multi-criteria decision analysis), off one weighted scoring engine. Use for panel scoring set-up, scoring grids, evaluation workbooks, or any weighted multi-criteria appraisal. Distinct from tor-procurement (writes and grades the ToR), procurement-offer-review (analyses one incoming offer), donor-proposal-scoring (proposals IPPF sends out rather than receives), and accreditation-desk-review (MAs against IPPF standards).

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.

Design MEL/SRHR indicators to IPPF/UNFPA/UNAIDS publication standard. Use when Ane asks for "indicators", "KPIs", "results framework", "M&E indicators", "measurement framework", or equivalent. Enforces WHO (2010) WHO/RHR/10.12 standards plus Washington Group disaggregation, applies Tier 1/2/3 integrity markers, defines measurement mechanisms, and flags data gaps. Distinguishes output, outcome, and impact indicators precisely.

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.

Work on a Word deliverable inside the .docx itself: read and engage reviewer comments, insert review findings as margin comments authored as Ane into her working copy, apply real Word tracked changes a counterparty can accept or reject, make scoped formatting-preserving fixes via Word COM, or add a branded glossary and verified source annex. Use when Ane hands over a commented or hand-edited .docx for review, markup, or in-file edits. Distinct from ane-voice (prose voice only, never touches the file), localise (target-language work), check-deliverable (read-only QA verdict), meeting-notes marginalia (reads comments, never writes them), and tor-procurement (builds ToRs).

Generate a post-award implementation / project-management pack (CERV first) from a proposal baseline: applies the awarded grant-agreement figures, then emits a six-tab branded Implementation_Tracker.xlsx, a periodic technical-report scaffold, and a README. Use when Ane sets up project tracking after an award or runs '/implementation-pack --donor cerv'; donor is an argument so Gates/OSF/UN are later --donor values. Does not replace the finance system of record, portal reporting, or the authorising officer's sign-off. Distinct from /proposal (pre-award).

Design MEL/SRHR data-collection instruments to IPPF/UNFPA publication standard: questionnaires, survey instruments, KII/IDI interview guides, focus group (FGD) guides, observation checklists, and topic guides, with disaggregation, consent, and care-referral for sensitive content built in by construction. Use when Ane asks for the data-collection tool itself, closing the gap between indicator design and analysis. Does not define indicators (indicator-designer) or analyse data.

Produce and publish the next entry in Ane's personal AI Field Notes series on LinkedIn, end to end: topic proposal from the series ledger, a dictation interview, the article plus short post, the companion toolkit, then publication via claude-in-chrome after per-artifact approval. Runs on Ane's personal Wine & Slate brand, never IPPF branding. Use for any personal AI/MEL thought-leadership piece for LinkedIn. Not for IPPF-branded MA-facing learning products (learning-product), meeting notes (meeting-notes), voice audits of existing text (ane-voice), or LinkedIn profile edits.

Score a written grant proposal against the donor's published award criteria (SCORE mode), or compare that scoring against the donor's own Evaluation Summary Report (CALIBRATE mode). Use when Ane asks how a donor would score an application, whether it passes the threshold, where points would be lost, or hands over a call document plus a submitted proposal. Distinct from /proposal (writes proposals TO donors), procurement-offer-review and tor-procurement (the buying side, supplier offers against a ToR), accreditation-desk-review (MAs against IPPF standards), and check-deliverable (a QA gate, not a scored assessment).

Enforce the IPPF/UNFPA/UNAIDS evidence-and-rigour citation standard on MEL/SRHR output: injects current authoritative framework versions with author and year, flags outdated versions, applies the data-gap protocol, and places citations tier-aware (Tier 1 evidence-base line, Tier 2 inline, junior-MEL annotated). Use whenever Ane produces a theory of change, evaluation design, indicator set, donor report, or SRHR programme analysis. Do not use for non-MEL work.

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 <path>`. 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.

Review an incoming supplier technical or financial offer against an IPPF ToR, and keep the internal procurement pack consistent: offer-review, pack-propagate, and score-compare modes. Use when Ane checks an offer or revised offer against a ToR and her comments, propagates a decision or figure across the procurement/waiver pack, or compares her panel scores with the AI scores. Distinct from tor-procurement (drafts and grades the ToR itself, upstream of offers), /proposal (funding proposals TO donors), implementation-pack (post-award tracking), and accreditation-desk-review (MA compliance).

Conduct an IPPF Member Association (MA) accreditation desk review against the IPPF EN Internal Guidelines Accreditation Review (Cycle 4 / ARV cycle, where ARV means Accreditation Review, NOT antiretroviral). Use when Ane assesses an MA self-assessment against accreditation principles, standards or checks, writes reviewer conclusions in the desk-review Word template, or prepares pre-interview document requests and interview questions, including on a bare principle, standard or check number ('do principle 8'). Not for literature reviews (evidence-synthesis), interview or survey guides (instrument-designer), indicator design (indicator-designer), or any clinical antiretroviral 'ARV' question.

Prepare the first working session with a contracted supplier against a fixed day budget: reads the agreed offer or contract, then builds a branded session workbook whose day-budget sheet totals against the contracted ceiling, an internal prep brief, a supplier agenda and a post-session note template. Use for kick-off or delivery-session prep with a supplier, or day-budget splits over a signed offer. Distinct from tor-procurement (writes the ToR), selection-toolkit (panel scoring), procurement-offer-review (one incoming offer), implementation-pack (donor grants IPPF receives), and meeting-notes mode 2 (1-1 prep).