Clews Consultancy

Showing 25 of 73 from the live probe.

• Cc Status

  CC server health and orientation. Returns server version, public mode, root availability, and warnings. Set include_profile=True to see the active profile and enforcement status. Set include_domains=True to see the list of enabled domains.

• Cc Capabilities

  What the CC server can do, by domain. Returns allowed operations, example tasks, output formats, and phase availability. Pass domain= to filter to one domain: dxl | memory | source | standards | dependencies Omit domain to list all domains.

• Cc Profile

  Manage the active internal profile. action: list — list all defined profiles with purpose and risk level get — show the current (or named) profile set — request a profile change (acknowledged; not enforced until Phase 2) profile_name: name of the profile to get or set (ignored for list). Common profiles: general_read, dxl_read, source_review, memory, standards.

• Cc Locate Capability

  Locate an MCP capability by intent/name. This is an orientation helper for AI-owned maintenance. It answers: - does the capability appear to exist? - where is it in source or memory? - is it directly exposed through the public facade? - what functions/files look relevant? It does not execute the located capability. Use this before concluding that something like the linter, class map, or code-wake protocol does not exist.

• Cc Search

  Search across CC knowledge domains. domain: dxl | memory | source | standards | dependencies query: search string, or empty / * to list all kind: optional sub-type filter (domain-specific) For dependencies: design_element | script_library | routine | unresolved_call path_prefix: restrict search to a sub-folder (relative path) modified_after: ISO date string e.g. 2026-01-01 to filter by modification date max_results: 1-100, default 20 profile: optional profile name (stateless; not enf

• Cc Read

  Read a CC knowledge artefact. Provide either: ref=<ref from cc_search> or: domain=<domain> + path=<relative path> output_format: full — complete content up to max_chars (default) summary — bounded structural preview max_chars: maximum characters to return (default 20000, max 100000) profile: optional profile name (stateless; not enforced in Phase 2) Returns content, metadata, truncation status, and suggested next actions.

• Cc Capability Discover

  Discover validated business capabilities behind the permanent gateway. Registry manifests are loaded dynamically on each invocation. Adding a capability therefore does not add another public MCP tool or require a ChatGPT MCP catalogue refresh. This surface is discovery-only.

• Cc Capability Read

  Invoke one registry-approved READ capability. The capability manifest fixes its adapter, authority profile and target bindings. Callers cannot provide an NSF, formula, executable, function, target or authority override through this gateway.

• Cc Capability Action

  Invoke one registry-approved WRITE capability. This surface rejects READ capabilities and dispatches only to an adapter fixed by a validated manifest. There is no generic Domino, LotusScript, NSF, formula, filesystem or authority execution parameter.

• Cc Task Authoring Reference

  Return current canonical CC Task authoring vocabulary. Values, schema identifiers, required fields, authority classes, provider-profile enums, work-value reasons, impact classes and execution modes are derived from the same TaskAuthoring Python contracts used by canonical admission. The route is read-only and performs no admission.

• Cc Task Authoring Preflight

  Read-only preflight for a candidate or proposal-shaped payload. Uses the current canonical TaskAuthoring validators and deterministic provider-profile derivation before admission. It never writes a task, queue, claim, run, admission receipt or other lifecycle state.

• Apply one bounded CC Task prompt micro-maintenance patch

  Evaluate or apply one governed CC Task prompt micro-maintenance patch. The route is deliberately narrower than the ordinary memory-write lane: only the canonical CC Task prompt and shared task-authoring instruction are eligible; patches must be literal, small, expected-hash gated, friction-backed and non-semantic. Authority, admission, queue/provider, eligibility, target/mutation, safety and CC Watch-authority changes are rejected to normal Prompt Library Governance. execute=False is preflight.

• Manage one exact known CC task

  Manage one exact canonical CC task without broad search. operation: STATUS — one exact canonical task-record read only. AMEND_REQUIREMENT — add/update one bounded requirement using the existing QueueControl CAS writer; expected task revision/hash and authority_ref are mandatory. SET_PRIORITY — update canonical task priority through the same CAS writer; expected version and authority required. REISSUE

• Cc Ai Friction Record

  Record one bounded AI/MCP friction observation for later F05 review. This is append-only operational evidence, not a task, repair or authority grant. Repeated identical observations within a short window are suppressed. Use it when the current work exposes avoidable reasoning, context, tool-call or workflow burden that would be useful to optimise.

• Cc Ai Friction Recent

  Read recent bounded AI/MCP friction observations for review or clustering.

• Cc Chatgpt Authoring Origin Capture

  Capture the exact ChatGPT conversation that authored a CC TASK request. Supply the exact current owner CC TASK message text and one stable authoring invocation ID. The tool searches synced ChatGPT history on the already-running managed Brave session, proves an exact user-message match, reads the settled canonical location.href, and returns validated authoring_context.chatgpt_ui provenance. It never sends a message, selects a provider, writes task/queue state, or guesses a project from titles or

• Cc Watch Latest State

  Read the deterministic CC Watch latest-state pointer. This is a read-only, bounded front door. It validates schema, integrity and freshness, returns current state before recent change, and expands at most two evidence records explicitly named by the pointer. It never crawls the queue tree or mutates Watch, task, claim, run or Domino state.

• Cc Task Friction Recent

  Read recent structured TASK_FRICTION evidence from the canonical outcome ledger. Recurrence counts are exact evidence-fingerprint counts over the same Task Outcome Ledger lineage. This route is read-only: it cannot write the Watch frustrations register, remediate anything, admit work or mutate queue/lifecycle state.

• Cc Management Action Review

  Review one exact immutable CC management-action proposal. The proposal ID and hash must match. The result shows the bounded task payload, current state and exact confirmation command only when this is an executable ISSUE_WORK draft with 1-5 tasks. This tool never mutates task, queue, proposal or admission state.

• Confirm one bounded CC task bundle

  Admit the exact task bundle named by an explicit owner command. Call only after the owner explicitly supplies the exact CC ISSUE MANAGEMENT ACTION proposal command in the current MCP-backed conversation. Set confirmation=true only for that explicit act. The operation revalidates current state, admits the whole 1-5 task bundle atomically, is replay-idempotent, and never dispatches work in the same cycle. DECIDE/zero-task proposals cannot use this tool.

• Cc Dxl Runtime Status

  Report active and recent DXL writeback sessions and output-file state. This reads the authoritative cc_ai_control runtime session tree rather than the optional legacy Pront transcript folder. Supply session_id to inspect one exact session. stale_after_seconds marks non-terminal sessions whose runtime evidence has not changed within the threshold.

• Cc Dxl Runtime Find Session

  Find DXL runtime sessions by identity, task, status, run, command, or output text.

• Cc Dxl Runtime Read Output

  Read a bounded window from an approved DXL runtime output file. Supported session outputs include status, command, active_manifest, worker_transcript, workbench_transcript, command_log, error_log, execution_proof and result. The current shared control outputs use the same names prefixed with current_. A stable ref returned by this tool may be supplied through path_or_ref on a later call.

• Cc Dxl Runtime Validate Contract

  Validate runtime-output identity across worker, workbench, status and command log. Contract v2 requires one shared session_id, command_id and run_id, explicit UTC timestamps, current actions, progress evidence and a consistent terminal result. Older sessions are reported as legacy_incomplete rather than being rewritten into fictional compliance.

• Cc Pront Transcript Status

  Report configured Pront transcript roots and latest visible files.