Skills synthese-multi-llm
📦
synthese-multi-llm
Content revision r1 High Risk 🔑 Env variables⚡ Contains scripts⚙️ External commands🌐 Network access📁 Filesystem access
Synthèse Multi-LLM avec conseil d’experts
Produit des synthèses robustes et traçables en faisant analyser une même source par plusieurs modèles aux rôles complémentaires, puis en consolidant leurs accords, divergences et points de vigilance.
Supports: Claude Codex Code(CC)
Stop for confirmation before installing.
Review the plan and obtain explicit user consent before changing files.
Install with my Agent
Copy this request to your Agent. It includes the canonical Skill page and manifest.
Agent request
Review the Skillstore skill "synthese-multi-llm" from https://skillstore.io/skills/alexmacapple-synthese-multi-llm.md and its manifest at https://skillstore.io/api/skills/alexmacapple-synthese-multi-llm/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.Your Agent should still show its plan and request any confirmation required by the security policy.
Agent-readable resources
Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.
Test it
Using "synthese-multi-llm". Un rapport stratégique de 30 pages doit être résumé pour un comité avec les risques et divergences.
Expected outcome:
- Synthèse consolidée en 10-15 lignes orientée décision.
- Liste des points d’accord entre modèles.
- Divergences ou incertitudes explicitement signalées.
- Références aux passages sources et trail d’audit si activé.
Using "synthese-multi-llm". Une synthèse existante doit être contrôlée contre son document source.
Expected outcome:
Le skill produit une revue critique signalant les omissions, reformulations risquées, glissements de sens et zones où les modèles ne convergent pas, avec recommandations de correction.
Security Audit
High Riskv9 • 7/9/2026 Open versioned report
The skill orchestrates multiple LLM backends and includes Python scripts, shell wrappers, network calls, environment-based API credentials, subprocess execution, and local audit/cache file writes. Static analysis produced many high-severity blocker matches, but most weak-cryptography and Markdown backtick alerts are false positives caused by French documentation text, PRD references, tables, and formatting. Confirmed risks are operational rather than obviously malicious: external CLI/API invocation, HTTP requests, environment secret handling, and filesystem writes. No net-new prompt-injection or malicious semantic finding was identified from the available static evidence.
21
Files scanned
8,591
Lines analyzed
82
Review items
0
False positives ignored
Confirmed security concerns (4)
High
Hex-encoded characters
"\x00", # Null byte
The source contains hex-style escape sequences or encoded characters; this is confirmed as encoded content, though not necessarily malicious by itself.
High
Hex-encoded characters
result = result.replace("\x00", "")
The source contains hex-style escape sequences or encoded characters; this is confirmed as encoded content, though not necessarily malicious by itself.
High
Hex-encoded characters
result = text.replace("\x00", "")
The source contains hex-style escape sequences or encoded characters; this is confirmed as encoded content, though not necessarily malicious by itself.
High
Hex-encoded characters
result = re.sub(r'[\x00-\x08\x0b\x0c\x0e-\x1f]', '', result)
The source contains hex-style escape sequences or encoded characters; this is confirmed as encoded content, though not necessarily malicious by itself.
Capability review items (82)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
Generic API/secret keys
| `ANTHROPIC_API_KEY` | Clé API pour backend Anthropic direct |
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
### "ANTHROPIC_API_KEY non défini"
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
1. Exporter la variable : `export ANTHROPIC_API_KEY="sk-ant-..."`
The documentation includes a placeholder API key format; it is not a real secret but confirms API-key handling guidance.
High
Generic API/secret keys
1. Vérifier la clé : `echo $ANTHROPIC_API_KEY | head -c 20`
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
- ANTHROPIC_API_KEY: Clé API (env var ou config)
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
- ANTHROPIC_API_KEY env var
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
- ou api_key dans le constructeur
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
api_key: Optional[str] = None,
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
api_key: Clé API (ou ANTHROPIC_API_KEY env var)
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
self._api_key = api_key
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
def api_key(self) -> Optional[str]:
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
return self._api_key or os.getenv("ANTHROPIC_API_KEY")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
- ANTHROPIC_API_KEY défini
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
if not self.api_key:
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
export ANTHROPIC_API_KEY="sk-ant-..."
The documentation includes a placeholder API key format; it is not a real secret but confirms API-key handling guidance.
High
Generic API/secret keys
api_key: "sk-ant-..."
The documentation includes a placeholder API key format; it is not a real secret but confirms API-key handling guidance.
High
Generic API/secret keys
self._client = anthropic.Anthropic(api_key=self.api_key)
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
api_key = os.getenv("ANTHROPIC_API_KEY")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
if not api_key:
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
return False, "ANTHROPIC_API_KEY non défini"
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
masked_key = api_key[:10] + "..." + api_key[-4:] if len(api_key) > 14 else "***"
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Hidden file in home directory
2. ~/.synthese-council/config.yaml (global utilisateur)
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
High
Hidden file in home directory
2. ~/.synthese-council/config.yaml - global utilisateur
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
High
Python subprocess.run
result = subprocess.run(
The code calls subprocess.run, confirming external process execution behavior that should be reviewed for argument safety.
High
Python subprocess.run
result = subprocess.run(
The code calls subprocess.run, confirming external process execution behavior that should be reviewed for argument safety.
High
Generic API/secret keys
" 1. Définir ANTHROPIC_API_KEY pour utiliser l'API Anthropic\n"
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
log("║ 2. Ou définir ANTHROPIC_API_KEY (fallback) ║")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
# - ANTHROPIC_API_KEY défini
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
if [ -z "$ANTHROPIC_API_KEY" ]; then
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
echo "Erreur: ANTHROPIC_API_KEY non défini" >&2
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
-H "x-api-key: $ANTHROPIC_API_KEY" \
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
sudo privilege escalation
sudo apt install curl
The documentation includes a sudo installation command; it is user-facing install guidance rather than automatic privilege escalation.
High
Generic API/secret keys
| `claude_wrapper.sh` | Claude (API) | `ANTHROPIC_API_KEY`, curl |
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
High
Generic API/secret keys
export ANTHROPIC_API_KEY="sk-ant-..."
The documentation includes a placeholder API key format; it is not a real secret but confirms API-key handling guidance.
High
Generic API/secret keys
| `ANTHROPIC_API_KEY` | Clé API Anthropic | - |
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
Medium
Unix shell invocation
#!/bin/bash
The file contains a shell shebang or shell script example, confirming shell execution content.
Medium
Hidden file access
2. ~/.synthese-council/config.yaml (global utilisateur)
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
Medium
Hidden file access
2. ~/.synthese-council/config.yaml - global utilisateur
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
Medium
Shell command substitution
"$(", # Command substitution
The shell wrapper uses command substitution to process prompts and capture curl responses, confirming shell execution behavior.
Medium
Shell command substitution
r"\$\(.*\)", # $(...) command substitution
The shell wrapper uses command substitution to process prompts and capture curl responses, confirming shell execution behavior.
Medium
Python file write/append
with open(cache_path, "w", encoding="utf-8") as f:
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
Medium
Python file write/append
with open(trail_file, "w", encoding="utf-8") as f:
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
Medium
Python file write/append
with open(md_file, "w", encoding="utf-8") as f:
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
Medium
Shell command substitution
ESCAPED_PROMPT=$(echo "$PROMPT" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read().str
The shell wrapper uses command substitution to process prompts and capture curl responses, confirming shell execution behavior.
Medium
Shell command substitution
RESPONSE=$(curl -s -X POST "$API_URL" \
The shell wrapper uses command substitution to process prompts and capture curl responses, confirming shell execution behavior.
Medium
Unix shell invocation
#!/bin/bash
The file contains a shell shebang or shell script example, confirming shell execution content.
Medium
Shell command substitution
ESCAPED_PROMPT=$(echo "$PROMPT" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read().str
The shell wrapper uses command substitution to process prompts and capture curl responses, confirming shell execution behavior.
Medium
Shell command substitution
RESPONSE=$(curl -s -X POST "$HOST/api/generate" \
The shell wrapper uses command substitution to process prompts and capture curl responses, confirming shell execution behavior.
Medium
Unix shell invocation
#!/bin/bash
The file contains a shell shebang or shell script example, confirming shell execution content.
Medium
Unix shell invocation
#!/bin/bash
The file contains a shell shebang or shell script example, confirming shell execution content.
Low
Python getenv function
return self._api_key or os.getenv("ANTHROPIC_API_KEY")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
Low
Python getenv function
api_key = os.getenv("ANTHROPIC_API_KEY")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
Low
getenv function call
return self._api_key or os.getenv("ANTHROPIC_API_KEY")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
Low
getenv function call
api_key = os.getenv("ANTHROPIC_API_KEY")
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
Low
Python environment access
env = os.environ.copy()
The matched location reads, checks, documents, or transmits environment-based API credentials, confirming environment access risk.
Low
Configuration library
command=config.get("command", cli_name),
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
args=config.get("args", []),
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
env=config.get("env", {})
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Hardcoded URL
lines.append("# Voir documentation: https://github.com/...")
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Configuration library
custom_roles = config.get("expert_roles", {})
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
return config.get("default_cadrage", DEFAULT_CONFIG["default_cadrage"]).copy()
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
backend = config.get("convergence", {}).get("backend", "hybrid")
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Python HTTP libraries
requests.RequestException: Si erreur réseau
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Python HTTP libraries
response = requests.get(url, headers=headers, timeout=timeout)
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Hardcoded URL
"Documentation: https://github.com/your-repo/synthese-council#installation"
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Configuration library
self.retry_config = self.config.get("retry", DEFAULT_RETRY_CONFIG)
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
cache_config = self.config.get("cache", {})
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
enabled=cache_config.get("enabled", True),
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
default_ttl=cache_config.get("ttl", 3600),
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
cache_dir=Path(cache_config.get("directory", ".cache/synthese")) if cache_config.get("directory") el
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
weights = self.config.get("convergence_weights", None)
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
timeout=self.config.get("timeout", 300),
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
if council.config.get("enable_trail", True):
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
sources = config.get("_sources", ["default"])
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Configuration library
sources = config.get("_sources", ["default"])
The code reads runtime configuration values, confirming configurable behavior that can affect execution, cache, timeout, or output paths.
Low
Hardcoded URL
Inspiré de [Council](https://github.com/bacoco/Council-board-skill) par bacoco et du concept [LLM Co
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Hardcoded URL
API_URL="https://api.anthropic.com/v1/messages"
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Hardcoded URL
# - OLLAMA_HOST: URL du serveur (défaut: http://localhost:11434)
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Hardcoded URL
HOST="${OLLAMA_HOST:-http://localhost:11434}"
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Standard device file access
if ! curl -s --connect-timeout 2 "$HOST/api/tags" > /dev/null 2>&1; then
The matched location performs or documents filesystem access, including writes, hidden paths, or device-file redirection.
Low
Python HTTP libraries
response = requests.post(
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Low
Hardcoded URL
"https://api.example.com/generate",
The matched location contains HTTP client usage or a fixed URL, confirming network-capable behavior or network documentation.
Risk Factors
🔑 Env variables (50)
references/configuration.md:128 references/troubleshooting.md:64 references/troubleshooting.md:69 references/troubleshooting.md:75 scripts/backends/anthropic_backend.py:108 scripts/backends/anthropic_backend.py:314 scripts/backends/anthropic_backend.py:108 scripts/backends/anthropic_backend.py:314 scripts/backends/anthropic_backend.py:9 scripts/backends/anthropic_backend.py:64 scripts/backends/anthropic_backend.py:65 scripts/backends/anthropic_backend.py:77 scripts/backends/anthropic_backend.py:85 scripts/backends/anthropic_backend.py:89 scripts/backends/anthropic_backend.py:106 scripts/backends/anthropic_backend.py:108 scripts/backends/anthropic_backend.py:116 scripts/backends/anthropic_backend.py:121 scripts/backends/anthropic_backend.py:135 scripts/backends/anthropic_backend.py:140 scripts/backends/anthropic_backend.py:170 scripts/backends/anthropic_backend.py:314 scripts/backends/anthropic_backend.py:315 scripts/backends/anthropic_backend.py:316 scripts/backends/anthropic_backend.py:319 scripts/backends/cli_backend.py:151 scripts/backends/cli_backend.py:275 scripts/backends/cli_backend.py:276 scripts/backends/cli_backend.py:277 scripts/config.py:514 scripts/config.py:533 scripts/convergence.py:376 scripts/synthese.py:1692 scripts/synthese.py:1711 scripts/synthese.py:1713 scripts/synthese.py:1714 scripts/synthese.py:1715 scripts/synthese.py:2115 scripts/synthese.py:2616 scripts/synthese.py:3128 scripts/synthese.py:3449 scripts/synthese.py:3491 scripts/synthese.py:1159 scripts/synthese.py:1765 wrappers/claude_wrapper.sh:9 wrappers/claude_wrapper.sh:23 wrappers/claude_wrapper.sh:24 wrappers/claude_wrapper.sh:42 wrappers/README.md:16 wrappers/README.md:45
⚡ Contains scripts (9)
⚙️ External commands (45)
scripts/backends/cli_backend.py:109 scripts/sanitize.py:27-42 scripts/sanitize.py:42 scripts/sanitize.py:180 scripts/sanitize.py:28-41 scripts/sanitize.py:41 scripts/sanitize.py:27-42 scripts/synthese.py:1229 scripts/synthese.py:1317 scripts/synthese.py:2851 scripts/synthese.py:2890 SKILL.md:20-32 SKILL.md:32-36 SKILL.md:36-38 SKILL.md:38-54 SKILL.md:54-70 SKILL.md:70-76 SKILL.md:76-77 SKILL.md:77-78 SKILL.md:78-79 SKILL.md:79-80 SKILL.md:80-81 SKILL.md:81-82 SKILL.md:82-83 SKILL.md:83-84 SKILL.md:84-98 SKILL.md:98-115 SKILL.md:115-146 SKILL.md:146-157 SKILL.md:157-159 SKILL.md:159-175 SKILL.md:175-185 SKILL.md:185-191 SKILL.md:191-195 SKILL.md:195-198 SKILL.md:198-202 SKILL.md:202-205 wrappers/claude_wrapper.sh:37 wrappers/claude_wrapper.sh:40-50 wrappers/claude_wrapper.sh:1 wrappers/ollama_wrapper.sh:37 wrappers/ollama_wrapper.sh:40-46 wrappers/ollama_wrapper.sh:1 wrappers/README.md:57 wrappers/README.md:142
🌐 Network access (10)
Audited by: claude View Audit History →
Share & cite this report
Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.
Copy report link
https://skillstore.io/skills/alexmacapple-synthese-multi-llm/audits/9?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown badge
[](https://skillstore.io/skills/alexmacapple-synthese-multi-llm?utm_source=security_passport_badge)HTML badge
<a href="https://skillstore.io/skills/alexmacapple-synthese-multi-llm?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/alexmacapple-synthese-multi-llm/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Embed card
<iframe src="https://skillstore.io/embed/skills/alexmacapple-synthese-multi-llm.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>Academic citations (APA · BibTeX · CFF)
APA citation
Alexmacapple. (2026). synthese-multi-llm security audit report (audit version 9) [Author version unspecified]. Skillstore. https://skillstore.io/skills/alexmacapple-synthese-multi-llm/audits/9BibTeX citation
@techreport{alexmacapple-alexmacapple-synthese-multi-llm-2026,
author = {Alexmacapple},
title = {synthese-multi-llm security audit report (audit version 9)},
institution = {Skillstore},
year = {2026},
number = {9},
url = {https://skillstore.io/skills/alexmacapple-synthese-multi-llm/audits/9},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "synthese-multi-llm security audit report (audit version 9)"
version: "unspecified"
type: report
authors:
- name: "Alexmacapple"
date-released: "2026-07-09"
url: "https://skillstore.io/skills/alexmacapple-synthese-multi-llm/audits/9"
identifiers:
- type: other
value: "skillstore:alexmacapple-synthese-multi-llm:audit:9"
description: "Skillstore immutable audit report identifier"
Skillstore Score
Why this score Evidence Confidence: Medium 68
Architecture
85
Maintainability
87
Content
70
Community
83
Spec Compliance
What You Can Build
Synthèse fiable de documents longs
Transformer un rapport, un dossier de recherche ou une note stratégique en synthèse structurée avec points d’accord, divergences et références aux passages sources.
Vérification critique d’une synthèse existante
Comparer une synthèse à son texte source pour repérer omissions, glissements sémantiques, surinterprétations ou désaccords entre modèles.
Préparation de notes adaptées à un public
Produire une restitution calibrée pour un destinataire précis, par exemple direction, équipe technique, comité pédagogique ou lectorat non spécialiste.
Try These Prompts
Synthèse exécutive traçable
Utilise synthese-multi-llm pour produire une synthèse exécutive de ce document. Cible: comité de direction. Finalité: décision rapide. Longueur: 12 lignes. Ton: clair et factuel. Mets en évidence les points d’accord, divergences et risques d’interprétation.
Analyse critique des glissements
Analyse cette synthèse par rapport au texte source avec le mode critique. Identifie les omissions, glissements sémantiques, extrapolations et points nécessitant vérification. Fournis les références aux passages sources quand elles sont disponibles.
Restitution pédagogique
Crée une synthèse pédagogique de ce contenu pour un public intermédiaire. Explique les concepts clés, conserve les nuances importantes et signale les désaccords entre modèles sous forme de points de vigilance.
Comparaison de sources multiples
À partir de ces sources, produis une synthèse multi-LLM comparative. Sépare les faits établis, les interprétations, les contradictions entre sources et les questions ouvertes à investiguer.
Best Practices
- Fournir des sources propres, complètes et clairement séparées pour faciliter les références et la comparaison.
- Définir le destinataire, la finalité, la longueur et le niveau d’expertise avant de lancer la synthèse.
- Relire les divergences et le trail d’audit avant d’utiliser la synthèse dans un contexte sensible.
Avoid
- Utiliser la sortie comme vérité définitive sans relecture humaine ni vérification des sources.
- Lancer des wrappers ou appels réseau avec des clés API sur des documents non fiables sans comprendre les flux de données.
- Demander une synthèse très courte pour un sujet fortement nuancé sans accepter de section sur les incertitudes.
Frequently Asked Questions
Quels modèles ce skill peut-il utiliser ?
Il est conçu pour orchestrer plusieurs backends comme Claude, Gemini, Codex ou des wrappers compatibles, selon ce qui est installé et configuré localement.
Le skill envoie-t-il des données à des services externes ?
Oui, il peut effectuer des appels réseau vers des fournisseurs de modèles ou des URL sources selon la configuration et les options utilisées.
Faut-il des clés API ?
Certains backends exigent des clés API, notamment via des variables d’environnement comme ANTHROPIC_API_KEY. Les backends locaux peuvent avoir d’autres prérequis.
Qu’apporte le multi-LLM par rapport à une synthèse classique ?
Le croisement de rôles et de modèles aide à faire apparaître les accords, divergences, omissions possibles et points de vigilance, au lieu de produire une seule perspective.
Les résultats sont-ils entièrement fiables ?
Non. La méthode améliore la robustesse et la traçabilité, mais une relecture humaine reste nécessaire pour les décisions importantes ou les documents sensibles.
Quels fichiers le skill peut-il générer ?
Selon la configuration, il peut écrire des caches, des trails d’audit JSON et des sorties Markdown dans les répertoires prévus par le projet.
Developer Details
Author
AlexmacappleLicense
MIT
Skillstore revision
r1
Version notice
The author did not declare a version.
Ref
7db9b9f06e0ab79c575b58bc48c4d8dc9849f424
Maintenance freshness
7/23/2026
Usage
7 downloads · 184 views