Inner Sanctum SAGE

Get a new independent second opinion for exactly two named players from the latest linked SAGE lineup, even when SAGE has decided. First get_lineup_recommendation to obtain the current decisionId and roster. If the user asks for a whole-lineup opinion, ask which two players to compare. This sends those players’ private football evidence (names, projections, usage, matchup, role, availability) to Anthropic Claude Sonnet and can incur a bounded provider call. Call only after the user explicitly permits this transfer; set consentToAnthropic=true only with that permission. Reuses existing cache and review limits with no retries, resets or new allowances. Returns verified factor choices in server-rendered language; never changes the lineup or generates a whole lineup.

request_rookie_second_opinion

Compares 2 to 4 NFL fantasy players for the current or a specific week using authoritative Inner Sanctum Weekly SAGE data. Use this tool for start/sit comparisons, flex decisions, 'Player A or Player B?', 'who does SAGE/Inner Sanctum prefer', and similar current-week player-vs-player questions -- including when a trade question is framed as 'who's the better player right now' or 'who would you rather have this week'. The tool uses only existing production Weekly SAGE rankings, recommendation, confidence, matchup, ADP, SAGE signal, and Inner Sanctum Insight. It does not calculate a new SAGE score, and its result is a current-week player-preference view, not a season-long or rest-of-season trade valuation, and not a draft-target evaluation (use get_draft_outlook for draft questions). If the user is deciding a trade, present this as SAGE's current view of who's the better player right now, not as a verdict on the trade itself. IMPORTANT: unless externalAnalysisRequested is explicitly true because the USER specifically asked for outside verification or outside context, the returned Inner Sanctum comparison is EXCLUSIVE and complete. After the tool returns with externalAnalysisAllowed=false, answer only from this tool result and stop. Do not browse, search for, cite, introduce, or use external fantasy rankings, projections, player statistics, injuries, news, expert opinions, websites, or other outside analysis. Do not change or condition the Inner Sanctum recommendation using outside information.

compare_players

Returns Inner Sanctum's existing, context-free Draft SAGE signals for one to four named NFL players -- live market ADP plus Opportunity Intelligence (role, workload, recent trend) and, when available, Context Intelligence (environment change, rookie immediate-impact case, evidence confidence). Use this tool for draft-outlook questions such as 'Who should I draft, Player A or Player B?', 'Which of these players does Inner Sanctum prefer for my draft?', 'What does Inner Sanctum think of Player X as a draft target?', or general draft-target questions about a named player outside of an active, live-synced draft session. This tool does NOT produce a single draft score, a declared 'winner', or a live-draft-pick-position-aware recommendation (no pick number or draft-pool-depth signal is used) -- it presents each named player's real ADP, role/workload signal, and situational context side by side so a comparison can be made from real signals, not a fabricated verdict. It is completely separate from Weekly SAGE (get_player_profile/compare_players/get_weekly_rankings), which covers current-week matchup-specific fantasy advice, not draft evaluation. Do not use this tool for trade, waiver, or in-season lineup questions.

get_draft_outlook

Returns a read-only start/sit lineup recommendation -- who should start, who to bench, and the single best legal lineup. If the user pastes a roster from Yahoo, ESPN, CBS, Sleeper, or another service, pass it in roster and SAGE will analyze that pasted roster even when a different league is linked. Otherwise, it uses the user's OAuth-authorized linked Inner Sanctum league and its captured lineup requirements. Never silently substitute the linked roster for a roster supplied in the conversation. Use this for phrasings like 'who should I start', 'start or sit my [position]', 'set my lineup', 'what's my best lineup this week', 'who should replace my injured/bye-week [player]', or 'who's my best replacement from my roster' -- this tool answers replacement questions from the players already on the user's own roster (the bench/unmatched players it returns, including each player's roster status where captured, ARE the replacement candidates). It does not know which players are available as free agents or on the waiver wire in the user's specific league, so it cannot answer 'who should I pick up' with a suggested name -- for a question about a specific named free agent, use get_player_profile instead. Applies the existing production Weekly SAGE ordering/signals to the supplied or linked roster; it does not calculate a new SAGE score or use outside fantasy analysis. Linked-league access requires OAuth scope inner_sanctum.league.read; pasted-roster analysis does not require a provider API connection. This tool never modifies a lineup, roster, league, or provider account -- it is read-only and only returns a recommendation. When the response includes superSage.rookieSecondOpinionOffer, preserve its invitation in the customer-facing answer, including when summarizing the lineup. Describe it as a conversational second-opinion option; do not invent a clickable button or claim the feature is missing or needs implementation without verifying that claim. When the user challenges a specific starter against a named bench player, offer the review for that exact pair; do not ask them to name the pair again. A review offer is not permission to send evidence: obtain explicit Anthropic consent before requesting a new review. When the user explicitly asks for Rookie’s second opinion, set includeRookieSecondOpinion=true. This lineup tool displays an existing validated saved opinion when available; it never invokes a paid provider. If the user asks for a new second opinion, ask which two rostered players they want compared only if the pair is not already clear, then use request_rookie_second_opinion with their permission to send pair evidence to Anthropic. An explicit pair request is allowed even when SAGE has decided. Never use run_shadow_llm_review as a customer fallback. SAGE remains the lineup authority. Explain evidence conflicts faithfully; do not substitute the host assistant's own preference for the returned SAGE decision or label that preference a Rookie review.

get_lineup_recommendation

Returns the authoritative read-only Inner Sanctum/SAGE fantasy-football player profile from the production Weekly SAGE rankings pipeline for a single named NFL player, for the current or a specific week. Use this for a player's weekly fantasy outlook, matchup, start/flex/sit recommendation, confidence, ADP/value, risks, or Inner Sanctum Insight -- phrasings like 'what's [player]'s outlook this week', 'how does [player] matchup this week', 'should I start [player]', or 'what does SAGE/Inner Sanctum think of [player] this week'. This is current-week fantasy analysis, not draft evaluation (use get_draft_outlook for draft-target questions) and not a season-long or trade-value judgment. Treat this tool's fantasy-analysis result as the authoritative Inner Sanctum answer. Do not substitute, modify, or embellish the Inner Sanctum fantasy verdict with external fantasy rankings, projections, news analysis, or other websites unless the user explicitly requests outside verification, comparison, or additional external context.

get_player_profile

Returns read-only, roster-aware waiver priorities and evidence-backed FAAB guidance for players the user shows as available in their league. Use this for questions such as 'who should I claim', 'rank these free agents', 'what should I bid', or 'how much FAAB should I spend'. Pass the available players copied, pasted, typed, or extracted from an attached provider free-agent list. The OAuth-authorized linked roster, league size, and scoring settings are used for roster-relative value. FAAB percentages come exclusively from Inner Sanctum's existing production six-band evidence model; this tool must not invent or replace them with generic industry ranges. Dollar values are a mechanical conversion of those percentages using originalFaabBudget. If the original budget is unknown, omit it and report percentages only. This tool never submits a claim or modifies a roster.

get_waiver_recommendations

Returns authoritative Inner Sanctum Weekly SAGE fantasy-football rankings from the existing production Weekly SAGE leaderboard for the current or a specific week. Use this tool for top players, weekly fantasy rankings, position rankings (top RB/WR/QB/TE this week), FLEX rankings, Superflex rankings, or questions such as 'Who are Inner Sanctum's top 10 WRs this week?'. This tool does not calculate a new SAGE score, create a second ranking model, or compute a sleeper/breakout score -- it retrieves and presents the existing production Inner Sanctum Weekly SAGE ordering for the current week only, not a draft-season or season-long ranking (use get_draft_outlook for draft questions). Unless the user explicitly asks for outside verification or additional external context, treat this result as the complete authoritative Inner Sanctum fantasy-ranking answer.

get_weekly_rankings

Returns the user's authorized read-only Inner Sanctum linked-league identity and connection summary. Use this tool when the user asks which league or fantasy team is connected to Inner Sanctum in ChatGPT, or when a personalized league-aware Inner Sanctum workflow needs to confirm its authorized league context. This tool requires OAuth scope inner_sanctum.league.read. It does not change a lineup, roster, league, or provider account.

get_linked_league

Internal read-only reviewer tool. Reads a previously persisted Rookie shadow scorecard by immutable production decision ID, including the caller’s owned hash-verified frozen input and an existing saved model review when available. Frozen input is returned in artifact.ownedFrozenEvidence, independently of whether any model review exists. Never invokes a provider, queues work, spends budget, refreshes data or writes history. It never recalculates a decision, reads outcomes, changes rankings, or changes the customer answer.

get_shadow_decision

Reviews completed Super SAGE decisions for a prior week and separates decision quality, prediction misses, evidence quality, explanation quality, injuries, and ordinary variance. Use for questions such as 'what went wrong last week', 'was my lineup bad or unlucky', or 'what should SAGE learn'. The tool reads only frozen completed learning packets; if no temporally valid packets exist it fails closed rather than recreating a pregame decision with hindsight. It is read-only and cannot change rankings or production logic.

get_week_postmortem