insat-pfe-report installable skill (Claude Code / opencode / Codex CLI)

A developer created an installable skill package for Claude Code, opencode, and Codex CLI that enables AI coding agents to generate INSAT end-of-studies (PFE) reports from a LaTeX template. The skill is discovered automatically when placed in a known directory with a SKILL.md file containing YAML frontmatter, and can be installed per project or globally.

Claude Code, opencode, and Codex CLI all now discover skills the same way: a folder containing a SKILL.md with YAML frontmatter name, description , dropped into a known directory. That's what this package is. Fastest path: one command cd insat-pfe-report this folder ./install.sh installs into the current project for all three tools or: ./install.sh --global installs into your home dir, available in every project This copies the skill into: .claude/skills/insat-pfe-report/ Claude Code .opencode/skills/insat-pfe-report/ opencode — also auto-detects .claude/skills/, so this is belt-and-suspenders .codex/skills/insat-pfe-report/ and .agents/skills/insat-pfe-report/ Codex CLI — the latter is the officially documented cross-tool path; installing both covers older and newer Codex builds Restart the tool / start a new session afterward so it re-scans for skills. Manual install, per tool Claude Code mkdir -p .claude/skills cp -r insat-pfe-report .claude/skills/ or globally: cp -r insat-pfe-report ~/.claude/skills/ opencode mkdir -p .opencode/skills cp -r insat-pfe-report .opencode/skills/ or globally: cp -r insat-pfe-report ~/.config/opencode/skills/ opencode also reads .claude/skills/ /SKILL.md automatically, so if you've already installed for Claude Code in the same repo, opencode picks it up with no extra step. Codex CLI mkdir -p .codex/skills or .agents/skills — Codex checks both, paths vary by version cp -r insat-pfe-report .codex/skills/ cp -r insat-pfe-report .agents/skills/ or globally: cp -r insat-pfe-report ~/.codex/skills/ Verifying it loaded Claude Code: ask "what skills do you have available?" or run claude skills list if your version supports it. opencode: the skill shows up in the <available skills list the agent sees; you can also check via the skill tool description. Codex CLI: run /skills or type $ to mention a skill by name to see it listed. If it doesn't show up: confirm the file is spelled exactly SKILL.md all caps , confirm the frontmatter has both name and description, and restart the session — none of these tools hot-reload skills mid-session. Using it Once installed, just ask the agent to work on the PFE report — e.g. "set up a new PFE report from the INSAT template" or "compile my report" from inside the project directory — and the description in SKILL.md is enough for the agent to pick it up on its own. You can also invoke it explicitly if your tool supports that Codex CLI: $insat-pfe-report . This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters Use this skill when the user wants to write, edit, or compile an INSAT end-of-studies PFE report using the ahmedsilinii/INSAT-PFE-Report-Template LaTeX template. Covers scaffolding a new report from the template, editing chapters/preliminary pages, regenerating the Word-based cover and last pages, and compiling to PDF locally pdflatex + biber without Overleaf. Trigger on mentions of "PFE report", "INSAT report template", "rapport de PFE", or a project directory containing this template's main.tex. license See upstream repo github.com/ahmedsilinii/INSAT-PFE-Report-Template for template license terms. Scripts in this skill are provided as-is. INSAT PFE Report — local build skill Turns github.com/ahmedsilinii/INSAT-PFE-Report-Template into something you can write and compile entirely on the command line — no Overleaf account needed. Overview The template is a LaTeX report-class document main.tex that: pulls chapters from chapters/ .tex general intro, chapter1–5, general conclusion pulls front/back matter from preliminary pages/ .tex dedication, acknowledgements, abstract, résumé, list of acronyms, appendices and bibliography from preliminary pages/bibliography.bib uses biblatex + biber with style=ieee set explicitly in main.tex: \usepackage style=ieee, backend=biber {biblatex} , tikz, pdfpages, pdflscape + eso-pic rotated landscape headers , listings, fancyhdr, titlesec, hyperref, plus table support tabularx, longtable, booktabs, xcolor table includes pages that are NOT written in LaTeX — they're Word docs exported to PDF and pulled in with \includepdf: PageDeGarde English.pdf cover — both pages of it are included: \includepdf{PageDeGarde English.pdf} then \includepdf pages=2 {PageDeGarde English.pdf} for an inner declaration page and LastPage.pdf back cover, single page , sourced from PageDeGarde English.docx and LastPage.docx respectively This has been verified end-to-end in a clean Ubuntu container: clone → install TeX Live + biber → 4-pass build pdflatex → biber → pdflatex → pdflatex, or latexmk -pdf as a one-shot → valid PDF. docx→PDF conversion for the cover/last page was verified with headless LibreOffice. First-time setup once per machine Run scripts/install-deps.sh. It installs, via apt: libreoffice headless docx→PDF conversion for the cover/last page — optional, skip with --no-libreoffice if the user only edits chapters This is a ~1–2GB install the first time. If the user is on a machine without apt macOS, etc. , tell them to install MacTeX or basictex + tlmgr install biblatex biber tikz pdfpages pdflscape listings fancyhdr titlesec enumitem csquotes and LibreOffice via brew install --cask libreoffice, then skip straight to compiling. Getting the template into a project Run scripts/fetch-template.sh <target-dir . It git clone --depth 1s the upstream repo into <target-dir and strips .git so it becomes the user's own project, not a fork tracking upstream. If the user already has the template cloned, skip this step and just cd into it. Compiling Run scripts/compile.sh <project-dir defaults to . . It runs, inside that directory: latexmk -pdf -interaction=nonstopmode -halt-on-error main.tex latexmk auto-detects that biblatex needs biber via main.bcf and reruns pdflatex as many times as needed to settle cross-references, the ToC, and citations — this replaces the manual pdflatex → biber → pdflatex → pdflatex sequence. Output is main.pdf in the project root. If latexmk isn't available for some reason, fall back to the manual sequence, run from the project directory: pdflatex -interaction=nonstopmode -halt-on-error main.tex biber main pdflatex -interaction=nonstopmode -halt-on-error main.tex pdflatex -interaction=nonstopmode -halt-on-error main.tex Two pdflatex passes after biber are required — one pass alone can leave the bibliography or ToC one revision stale. On a clean compile you'll see a handful of harmless pdfTeX warning ... PDF inclusion lines from the \includepdf calls and possibly a destination with the same identifier warning — these are not errors. Treat anything under LaTeX Error or Undefined control sequence in the log as the thing to actually fix. If the user only changed a .tex file no new citations, no cover page swap , a single pdflatex -interaction=nonstopmode main.tex is enough for a quick preview — just remember to do the full 4-step or latexmk pass before calling it done, or citations/toc will be stale. Editing content Chapters: edit chapters/chapter1.tex … chapter5.tex, general introduction.tex, general conclusion.tex. To add/remove a chapter, edit the \input{chapters/...} lines in main.tex directly — mirror the existing pattern. References: add BibTeX entries to preliminary pages/bibliography.bib, cite with \cite{key}. Style is IEEE via biblatex — no manual formatting needed. Images: put files under assets/ e.g. assets/Logos/ and \includegraphics them. Code listings: use the lstlisting environment already configured with line numbers, a light gray background, and syntax highlighting via the listings package . Editing the cover and last page no Word needed These two source files are DOCX documents exported to PDF; main.tex includes the PDFs, not the DOCX. PageDeGarde English.docx must stay a 2-page document cover + inner declaration page since main.tex pulls both pages in separately — deleting a page or adding one shifts what \includepdf pages=2 {...} grabs. LastPage.docx is a single page. Two ways to edit them without opening Word/Overleaf: Text edits only, keep the Word layout: edit PageDeGarde English.docx / LastPage.docx with any tool the user has LibreOffice Writer GUI, or programmatically with python-docx if it's a simple find-and-replace like a name or project title , then run scripts/convert-cover-pages.sh <project-dir to headless-convert both docx files back to PDF with the exact same filenames PageDeGarde English.pdf, LastPage.pdf that main.tex expects. Re-run the compile after. Full redesign: if the user wants a different look, they can design a new PDF Canva, LaTeX, whatever and drop it in under the same filename — main.tex doesn't care how the PDF was made. Never hand-edit the .pdf cover files directly; always regenerate from source and keep filenames identical, since \includepdf{PageDeGarde English.pdf} and \includepdf{LastPage.pdf} are literal filename references in main.tex. Common failure modes LaTeX Error: File 'preliminary pages/bibliography.bib' not found: run from the project root, not from inside chapters/ or preliminary pages/ — all paths in main.tex are relative to the root. Citations show as ? or bibliography is empty: biber didn't run, or ran before a .bcf was generated. Always do a pdflatex pass before biber, or just use latexmk -pdf which sequences this correctly on its own. biber: command not found: rerun scripts/install-deps.sh, or apt install biber / tlmgr install biber depending on the TeX distribution. Cover/last page missing or stale after a docx edit: the PDF wasn't regenerated, or was saved under a different filename — rerun scripts/convert-cover-pages.sh and check the two PDF filenames match exactly case-sensitive . Space in preliminary pages/ trips up a shell one-liner: always quote the path, e.g. "preliminary pages/bibliography.bib".