When the user asks "cosa ricordi di X?", "dammi tutto su Y", "chi è Z?", "cosa abbiamo detto di W?", "ricostruisci ciò che sai di [Capitalized name or known alias]", call hippoentityget(name) BEFORE answering. Triggers also on "what do you know about X", "tell me about Y", "facts about Z", "history of W". The tool returns entity+aliases+facts from a persistent SQLite KG (P2.a entity-centric KG, schema v4, Unicode-safe). Lookup is case-insensitive Unicode (Müller/MÜLLER/müller same match) and resolves aliases ("S. Tonegawa" → canonical "Tonegawa"), returning the list of factid linked to the entity. If entity is None, declare honestly "non ho memoria di X" — do NOT confabulate from training data. If entity exists but facts == [], declare "ho il nodo entity ma nessun fatto collegato" and suggest hippofactssearch or hipporecall. Zero LLM cost, sub-50ms on ~500 entity corpus, free in hosted mode. Backend SQLite WAL + UNIQUE INDEX on namenorm Python-normalized (NFC + str.lower, NOT SQLite ASCII-only LOWER). Entity-first recall — complements hippofactssearch (topic-based) and hipporecall (semantic episode) with a third modality, graph node lookup. ANTI-PATTERN: do NOT use for factual claim validation ("Tonegawa won Nobel in 2014") — that's hippovalidateclaim. Do NOT use for topic-based episode/dialog search — that's hipposearch or hipporecall.
EnglishANY mention of "memoria/memory/ricordi/ricordo/ricorda/ricordare/saved/stored" by the user MUST go to HippoAgent (hippo tools), NEVER to local file-system memory files (CLAUDE.md / MEMORY.md / .claude/projects/.../memory/). Auto-fire flow - call hippohealth first to confirm reachable, then route the user's intent to the right tool. "salva/save X" → hipporemember. "qual è il mio X / what's my X / cosa ricordi di X" → hippofactssearch (instant keyword) then fall back to hippofactsrecall (semantic). "cosa vedi nella memoria / what's in memory / dammi tutto" → hippofactslist + hippoepisodelist + hippostats. "abbiamo già fatto X / have we done X" → hipposearch (keyword on episodes) then hipporecall (semantic). "dimentica X" → hippoforget / hippofactforget. AVOID hipporuntask and hippoconsolidate inside Claude Code — they make EXTRA API calls to the configured Anthropic key (separate billing). The other 37 tools are FREE - they only read/write local SQLite, no API. Disable globally with env HIPPODISABLED=1.
EnglishBefore asserting any verifiable factual claim (year/number/attribution/citation/date), call hippovalidateclaim(claim) to check it against persistent memory. Triggers when about to say things like "X was born in YEAR", "Y won the Nobel in YEAR", "Z released VERSION in MONTH", "W said EXACTQUOTE", "this library/paper/repo is from ORG", or any other claim containing a Capitalized name + a specific year/number/version/org. The tool returns verdict ∈ {supported, contradicted, unknown} + an advice string in Italian (e.g. "in memoria 1987 (fact fid), NON 2014 — controlla prima di affermare"). If verdict=contradicted, STOP and re-read the evidencefacts; do NOT assert the claim verbatim. If verdict=unknown, hedge ("non ricordo con certezza", "credo ma verifica"). If verdict=supported, proceed. Zero LLM cost (deterministic, sub-100ms). Anti-confabulation gate — born from live confabulations 2026-05-14 (Tonegawa Nobel 1987→2014, Anthropic Skills Oct 2025→Oct 2026, LightRAG HKUDS→HKUST).
English
AGENT SKILLS ATLASIncludes Hippoagent Entity Recall, Hippoagent Memory, and Hippoagent Validate Claim.
Claude Code
Codex
Cursor
GitHub Copilot
Gemini CLISupported by the Atlas install prompt. Skill-specific requirements vary.
Pack stats
What’s included
Hippoagent Entity Recall
When the user asks "cosa ricordi di X?", "dammi tutto su Y", "chi è Z?", "cosa abbiamo detto di W?", "ricostruisci ciò che sai di [Capitalized name or known alias]", call hippoentityget(name) BEFORE answering. Triggers also on "what do you know about X", "tell me about Y", "facts about Z", "history of W". The tool returns entity+aliases+facts from a persistent SQLite KG (P2.a entity-centric KG, schema v4, Unicode-safe). Lookup is case-insensitive Unicode (Müller/MÜLLER/müller same match) and resolves aliases ("S. Tonegawa" → canonical "Tonegawa"), returning the list of factid linked to the entity. If entity is None, declare honestly "non ho memoria di X" — do NOT confabulate from training data. If entity exists but facts == [], declare "ho il nodo entity ma nessun fatto collegato" and suggest hippofactssearch or hipporecall. Zero LLM cost, sub-50ms on ~500 entity corpus, free in hosted mode. Backend SQLite WAL + UNIQUE INDEX on namenorm Python-normalized (NFC + str.lower, NOT SQLite ASCII-only LOWER). Entity-first recall — complements hippofactssearch (topic-based) and hipporecall (semantic episode) with a third modality, graph node lookup. ANTI-PATTERN: do NOT use for factual claim validation ("Tonegawa won Nobel in 2014") — that's hippovalidateclaim. Do NOT use for topic-based episode/dialog search — that's hipposearch or hipporecall.
Hippoagent Memory
ANY mention of "memoria/memory/ricordi/ricordo/ricorda/ricordare/saved/stored" by the user MUST go to HippoAgent (hippo tools), NEVER to local file-system memory files (CLAUDE.md / MEMORY.md / .claude/projects/.../memory/). Auto-fire flow - call hippohealth first to confirm reachable, then route the user's intent to the right tool. "salva/save X" → hipporemember. "qual è il mio X / what's my X / cosa ricordi di X" → hippofactssearch (instant keyword) then fall back to hippofactsrecall (semantic). "cosa vedi nella memoria / what's in memory / dammi tutto" → hippofactslist + hippoepisodelist + hippostats. "abbiamo già fatto X / have we done X" → hipposearch (keyword on episodes) then hipporecall (semantic). "dimentica X" → hippoforget / hippofactforget. AVOID hipporuntask and hippoconsolidate inside Claude Code — they make EXTRA API calls to the configured Anthropic key (separate billing). The other 37 tools are FREE - they only read/write local SQLite, no API. Disable globally with env HIPPODISABLED=1.
Hippoagent Validate Claim
Before asserting any verifiable factual claim (year/number/attribution/citation/date), call hippovalidateclaim(claim) to check it against persistent memory. Triggers when about to say things like "X was born in YEAR", "Y won the Nobel in YEAR", "Z released VERSION in MONTH", "W said EXACTQUOTE", "this library/paper/repo is from ORG", or any other claim containing a Capitalized name + a specific year/number/version/org. The tool returns verdict ∈ {supported, contradicted, unknown} + an advice string in Italian (e.g. "in memoria 1987 (fact fid), NON 2014 — controlla prima di affermare"). If verdict=contradicted, STOP and re-read the evidencefacts; do NOT assert the claim verbatim. If verdict=unknown, hedge ("non ricordo con certezza", "credo ma verifica"). If verdict=supported, proceed. Zero LLM cost (deterministic, sub-100ms). Anti-confabulation gate — born from live confabulations 2026-05-14 (Tonegawa Nobel 1987→2014, Anthropic Skills Oct 2025→Oct 2026, LightRAG HKUDS→HKUST).
Quick start
- Choose your skills. Select individual skills above, or choose the entire pack.
- Copy the install prompt. Use Copy install prompt above the list.
- Send it to your agent. Paste into your coding agent’s chat. Your agent reviews the source and guides the setup.
Try it with your agent
After installation, start with a request like:
Use the Hippoagent Entity Recall skill to help with my task.
Ask me for the context you need before getting started.Requirements & license
- A coding agent that can access your project files and install skills.
- Your agent checks each skill’s instructions, dependencies, and permissions during setup.
- Reported license for the skills shown: MIT. The author’s terms apply.