Capability, coverage, and claim map

Generated by scripts/capability_map.py; CI checks freshness. Use the JSON companion for agent consumption.

Result contract

Canonical promise: A legal result means only that the query's named checked scope passed. It is not a claim of global or complete legality.
Authority: srdcheck reports an advisory SRD-derived rule_result. The authorized DM, whether human or agent, owns table_decision; recording one never rewrites the rule result or implies state mutation.
Prose: The why field is explanatory and non-contractual. Consumers obtain the verdict schema identity from --schema, MCP outputSchema, or capabilities, then branch on verdict, exit_code, coverage_level, checked_scope, unchecked_scope, assumptions, facts, rule_result, table_decision, state_mutation, rule_ids, citations, and structured data instead of parsing why.
Vision: Target capabilities are labeled target and are never advertised as executable tools.

Exit-2 recovery contract

Machine metadata lives in data; agents branch on it instead of parsing why. An authorized agent-DM may resolve a DM-authority table ruling directly.

reason	recoverability	default next action	authority
`invalid-input`	`retry`	`repair-request`	`—`
`missing-fact`	`retry`	`provide-facts`	`—`
`unsupported-content`	`alternate-path`	`select-adapter`	`—`
`unmodeled-rule`	`alternate-path`	`use-other-capability`	`—`
`rules-ambiguous`	`authority`	`resolve-table-ruling`	`dm`
`gm-discretion`	`authority`	`resolve-table-ruling`	`dm`
`stale-state`	`conflict`	`reconcile-state`	`—`

Privacy-safe observability contract

srdcheck.observability/1.0 is out-of-band, disabled by default, and metadata-only.

Release identity

Engine 0.9.0; verdict schema 4.0; compatibility N/N-1 over 0.9, 0.8.

adapter	adapter version	data version	rules version	digest
`srd-5.1`	`0.1.0`	`0.1.0`	`0.0.0`	`39e4a7891480…`
`srd-5.2.1`	`0.2.1`	`0.2.1`	`0.2.1`	`fc9f0ccaa099…`
`toy-tictactoe`	`0.1.0`	`0.1.0`	`0.1.0`	`45746ba2683f…`

Shipped, executable today

A legal result below means only that the named checked scope passed.

tool	capability	checked scope	explicitly unchecked	evidence
`jurisdiction`	Check whether a name occurs in a loaded adapter registry and report every matching content category.	registry membership; all matching registry categories	mechanical legality; whether an unknown name is a typo, homebrew, or third-party content	`tests/test_kernel.py::test_jurisdiction_known_content`; `tests/test_kernel.py::test_jurisdiction_unknown_content`; `tests/test_kernel.py::test_jurisdiction_returns_all_categories_for_colliding_name`
`turn_plan`	Check a proposed turn against the modeled action-economy and movement budgets.	action, bonus-action, reaction, interaction, movement, and spell-slot-per-turn budgets; modeled condition effects on those budgets	whether a feature grants an action; targets, ranges, line of sight, cover, inventory, and bespoke feature prerequisites	`tests/test_turn_economy.py::test_two_bonus_actions_illegal`; `tests/test_enumeration.py::test_t5_consistency_sweep`
`turn_options`	Enumerate remaining modeled action-economy option kinds.	remaining budget option kinds; movement remaining and modeled stand/crawl costs	every legal creature action; feature-specific actions, targets, geometry, and tactics	`tests/test_enumeration.py::test_t5_consistency_sweep`
`reaction_available`	Check whether the modeled reaction budget is available.	reaction spent since turn start; Incapacitated and the Stunned, Paralyzed, Petrified, and Unconscious definitional embeds	whether a proposed trigger occurred; target and range legality	`tests/test_turn_economy.py::test_reaction_available_honors_incapacitated_and_embeds`; `tests/test_turn_economy.py::test_reaction_available_refuses_unknown_and_non_condition_names`
`roll_compose`	Compose caller-supplied Advantage and Disadvantage sources.	cancellation and non-stacking of supplied sources; reroll-one-die note	discovering sources from narrative or character state; rolling dice	`tests/test_roll_modifiers.py::test_two_adv_one_dis_is_straight`; `tests/test_roll_modifiers.py::test_stacked_advantage_still_two_dice`
`attack_modifiers`	Compose modeled attack modifiers from structured combatant facts.	documented condition modifiers; Prone distance behavior; condition-sensitive fact dependencies and category gates; ranged attacks in close combat with Incapacitated embeds	cover and line-of-sight geometry not supplied as facts; weapon or feature prerequisites	`tests/test_condition_fact_dependencies.py::test_attack_condition_fact_matrices`; `tests/test_condition_fact_dependencies.py::test_ranged_enemy_dependency_claim_evidence`; `tests/test_condition_fact_dependencies.py::test_condition_category_gate_claim_evidence`; `tests/test_condition_fact_dependencies.py::test_transport_parity_claim_evidence`
`mage_hand_use`	Check one proposed Mage Hand use against its modeled grants, prohibitions, weight, and range.	listed uses and prohibitions; 10-pound limit; 30-foot range for the proposal	fine-manipulation ambiguity; duration or range trigger tracking across events; narrative consequences	`tests/test_kernel.py::test_mage_hand_demo_goldens`
`event_apply`	Evaluate one supported event into a precondition-bound, idempotent transition proposal.	declared event validation and supported state transitions; full state-precondition and deterministic result hashes; stable transition identity and rule/ruling lineage	persisting state; generating events; advancing time autonomously; geometry; bespoke spell effects	`tests/test_ledger.py::test_lineage_chain_and_tamper_detection`; `tests/test_state_precondition_commits.py::test_proposals_and_result_hashes_are_deterministic_and_input_bound`
`transition_commit`	Verify a state-bound transition for an atomic caller-owned commit or idempotent retry.	state compare-and-swap precondition; deterministic proposal integrity; idempotent committed-result replay; typed stale-state conflicts	server-side persistence; choosing event order; distributed locking; Discord delivery guarantees	`tests/test_state_precondition_commits.py::test_commit_is_idempotent_without_server_side_state`; `tests/test_state_precondition_commits.py::test_concurrent_and_out_of_order_discord_events_conflict_then_reconcile`; `tests/test_state_precondition_commits.py::test_tampered_proposal_and_false_retry_fail_closed`
`creature_valid`	Classify whether a name is a registered SRD 5.2.1 creature.	creature registry membership; source citation for a match	creature action legality; third-party content; typo diagnosis	`tests/test_creature_queries.py::test_valid_creature_is_legal_and_cited`
`creature_stats`	Return registered Challenge Rating and XP facts for a creature.	name, Challenge Rating, XP, and source citation	stat-block actions and traits; encounter or turn legality	`tests/test_creature_queries.py::test_stats_returns_cr_xp_citation`
`encounter_xp_budget`	Compute the SRD encounter XP budget from supplied level, difficulty, and party size.	per-character and total XP budget arithmetic	encounter suitability; monster tactics; party composition beyond size and level	`tests/test_encounter_budget.py::test_issue_example`; `tests/test_encounter_budget.py::test_cited`
`save_check`	Compose or resolve a saving throw from caller-supplied facts and roll.	DC comparison; modeled condition modifiers and auto-fails; supplied d20 and modifiers	rolling dice; deriving sheet modifiers; effects of success or failure	`tests/test_combat_resolution.py::test_save_meets_or_exceeds_dc`; `tests/test_combat_resolution.py::test_paralyzed_petrified_stunned_unconscious_auto_fail_str_dex`
`check_make`	Compose or resolve an ability check from caller-supplied facts and roll.	DC comparison; modeled condition and social modifiers; condition-sensitive fact dependencies and category gates; supplied d20 and modifiers	rolling dice; choosing a DC; deriving modifiers; narrative outcome	`tests/test_condition_fact_dependencies.py::test_check_condition_fact_matrices`; `tests/test_condition_fact_dependencies.py::test_condition_category_gate_claim_evidence`; `tests/test_condition_fact_dependencies.py::test_transport_parity_claim_evidence`
`concentration_check`	Compute and optionally resolve the concentration save caused by supplied damage.	damage-derived DC; supplied Constitution modifier and d20 result	tracking concentration over time; detecting damage events; rolling dice	`tests/test_combat_resolution.py::test_concentration_dc_is_half_damage_floor_10_cap_30`; `tests/test_combat_resolution.py::test_concentration_resolves_when_roll_supplied`
`opportunity_attack_provoked`	Check provocation from caller-supplied movement and visibility facts.	supplied movement kind; supplied reach-exit and visibility facts on a potentially provoking voluntary path; Disengage, teleportation, and forced-movement exceptions	facts the caller omitted; computing geometry or reach exit; reaction availability; attack resolution	`tests/test_opportunity_attack.py::test_voluntary_movement_by_a_seen_creature_provokes`; `tests/test_opportunity_attack.py::test_disengage_teleport_and_forced_movement_do_not_provoke`; `tests/test_opportunity_attack.py::test_missing_positive_prerequisites_refuse_instead_of_defaulting_true`
`grapple_initiate`	Check Grapple or Shove initiation and compose its save DC formula.	caller-supplied attacker and target sizes; caller-supplied free-hand fact for Grapple; save DC formula and supplied formula blanks	prerequisite facts the caller omitted; escape contests; reach geometry; rolling or resolving the target save	`tests/test_grapple.py::test_target_more_than_one_size_larger_is_impossible`; `tests/test_grapple.py::test_grapple_requires_a_free_hand`; `tests/test_grapple.py::test_missing_prerequisite_facts_refuse_instead_of_defaulting_medium_or_free`; `tests/test_formula_blanks.py::TestGrappleBlanks::test_no_blanks_yields_formula_never_a_number`
`passive_perception`	Compute Passive Perception from a supplied modifier.	10 plus supplied Wisdom (Perception) modifier; refusal of the non-SRD plus-or-minus-five convention	deriving the modifier; Hide contests; line of sight and lighting	`tests/test_passive_perception.py::test_score_is_10_plus_modifier`; `tests/test_passive_perception.py::test_advantage_disadvantage_are_refused_as_non_srd`
`help_assist`	Check modeled prerequisites for the Help action.	caller-supplied ability-check proficiency fact; caller-supplied attack-assist five-foot fact; Advantage grant	facts the caller omitted; whether assistance is fictionally possible; geometry discovery; the authorized DM's table ruling	`tests/test_help_action.py::test_assist_ability_check_needs_the_relevant_proficiency`; `tests/test_help_action.py::test_assist_attack_roll_needs_an_enemy_within_5ft`
`spell_facts`	Return registered casting-time, range, component, duration, and concentration facts.	registered spell facts; fixed-duration expiry arithmetic on a caller clock	spell effects; whether casting is legal; slots, targets, and stateful expiry triggers	`tests/test_spell_facts.py::test_ward_regression`; `tests/test_spell_facts.py::test_mage_hand_not_concentration`
`feature_uses`	Return the modeled use-count formula for the closed feature set.	Divine Sense and Lay on Hands formulas; arithmetic when all blanks are supplied	other features; tables; whether a use remains in campaign state	`tests/test_spell_facts.py::test_divine_sense_blanks`; `tests/test_spell_facts.py::test_lay_on_hands_blanks`
`ttt_move`	Check one move against the bundled tutorial ruleset.	board shape, turn order, cell occupancy, and terminal state	strategy; choosing a move	`tests/test_toy_adapter.py::test_move_legality`
`ttt_options`	Enumerate legal cells under the bundled tutorial ruleset.	legal empty cells for the current player	strategy; ranking options	`tests/test_toy_adapter.py::test_enumerate_validate_consistency`
`table_evaluation`	Project one native rules query into table.evaluation/1.0 with machine-readable scope, caller references, and portable evidence.	deterministic projection of one named query; self-attested coverage and exact finding evidence; caller-owned session, entity, and correlation references	global action or table legality; table, encounter-state, or execution authority; protected-host attestation	`tests/test_table_evaluation.py`; `tests/test_mcp.py::test_table_evaluation_tool_is_opt_in_scoped_and_joinable`; `scripts/cold_artifact_smoke.py`

Target architecture — not shipped

These entries are roadmap intent, not executable capability claims.

target	intended capability	current truth
`batched-turn-orchestration`	Adjudicate a natural-language, multi-rule turn in one call.	Current tools accept structured, capability-specific calls; the caller parses intent and orchestrates them.
`creature-action-enumeration`	Enumerate every legal move for a creature from its full features and scene state.	turn.options enumerates modeled budget kinds, not creature-specific actions, targets, or tactics.
`spatial-geometry-cover-light`	Derive line of sight, lighting, cover, reach, and legal destinations from a map.	Current checks consume relevant geometry facts supplied by a caller or another rail.
`hide-lifecycle`	Adjudicate Hide preconditions, legal locations, and its future end triggers.	Current tools can resolve a caller-structured check and Passive Perception but do not implement Hide or its lifecycle.
`jump-distance`	Adjudicate and enumerate jump destinations.	No shipped query models jumping.
`duration-trigger-tracking`	Arm and resolve spell, condition, and feature triggers across future events.	spell.facts reports durations and event.apply folds only its declared event set; neither autonomously tracks future triggers.
`character-build-legality`	Validate level-up and character-build prerequisites.	No shipped query adjudicates builds or stamps level-ups.

FilesExpand file tree

capability-map.md

Latest commit

History