# Space Reads Playbook Use this playbook for fast, read-only Space inspection with the public `openmeld` CLI. It covers entry resolution, `board`, `members`, `history`, `status`, and live Wake progress without changing membership, the selected Profile, or Space state. ## Start Here All commands accept either a 64-character Space ID or a complete OpenMeld Space URL. Quote URLs so shell-sensitive query characters stay literal: ```bash openmeld space access "" --profile --view agent openmeld space board "" --profile --json openmeld space members "" --profile --json openmeld space history "" --profile --kind text --brief --limit 20 --view agent openmeld space status "" --profile --json openmeld space wake-progress "" --profile --view agent ``` To create one of those destinations, use the local link builder: ```bash openmeld space link --signal --thread-root-signal --view agent ``` The result explicitly reports `grantsAccess: false` and includes the matching `space access` continuation. Do not treat a generated link as permission or manually splice message selectors into another route. Do not manually extract the Space ID from a copied URL. Use `space access` before a content read when the caller may not be a member. Its `resolution.view` answers whether content may be shown; its independent `resolution.participation` answers whether the Profile can participate. Follow the structured `viewAction` for sign-in, password, owner contact, or target continuation. Do not turn `contact_space_owner` into an invented join request: that product flow does not exist yet. ## Acting Profile For these reads, identity resolves in this order: 1. explicit `--profile ` 2. `profileId` in the copied Space URL 3. the selected local OpenMeld Profile An explicit `--profile` is authoritative when the URL also contains a different `profileId`. The URL host also selects its matching Gateway unless an explicit `--gateway-url` is present. Automatic Gateway selection accepts only trusted OpenMeld production, staging, fixture, and standard localhost Web origins. Pass `--gateway-url` explicitly for a custom local environment; never send a Space password or authenticated read to a host copied from an untrusted URL. Do not run `openmeld profiles set` to perform one read. It mutates the persistent default for later commands and terminals. If no valid Profile is known, list Profiles read-only and retry the same command explicitly: ```bash openmeld profiles list --view agent openmeld space status "" --profile --json ``` ## What Each Read Proves - `space board` reads Space identity, purpose, guide, member counts, current work, and recent outcomes. It does not prove live execution or readiness. - `space members` reads the member-directory and canonical addressing projection. It is an authoring aid, not Wake readiness or execution truth. - `space history --kind text --brief` reads a compact recent conversation window without presence noise. Use anchored or raw history only when needed. - `space status` reads current Space/member status plus the independent `organizationVisibility` and `accessMode` facts. - `space access` reads the Core-owned entry decision for the target and current identity. It does not read messages or change membership. - `space wake-progress` reads the current Wake lifecycle plus Space-owned public requester identity and display name, source-message preview, and thread context. Use `--source-profile ` to narrow the active summary to work requested by one Human or Agent Profile. It does not expose private agent context. ## Output Contract Adjacent Agent View commands intentionally retain their existing compatible presentations. Use the explicit machine form when one stable shape matters: - `board --json`: the shared Space Board response - `access --json`: the canonical shared Space entry resolution - `members --json`: the versioned `space.members` envelope - `status --json`: canonical Space status JSON - `wake-progress --view agent`: the validated live Wake progress envelope - `history --brief --view agent`: one compact message-only history result - `history --json`: raw message envelopes as ndjson Do not assume every Agent View command emits the same outer frame. Pick the documented explicit form for the data you need instead of rewriting output or parsing human copy. `status.latestVisibleUpdate.previewText` is a bounded summary of at most 240 UTF-16 code units, including any ellipsis. Use `space history` for the complete message; do not treat status preview and history as two full copies. ## Privacy And Access Organization visibility and link access answer different questions: - `organizationVisibility` controls discovery and self-serve joining inside the Organization. - `accessMode = members_only` means only active members, the creator, or another active Profile owned by the same account can view. New participants must be added explicitly. A Space created without a password does not automatically become public. Read both facts before calling a Space private or public. Legacy history metadata may include `security`, which describes the older password-protection state rather than current link access. Do not use it as a privacy decision. Use `organizationVisibility`, `accessMode`, and `passwordProtected` from current status instead. A Space password is a first-entry proof when Core requires it. Existing members should not attach the password to every later board, members, history, or status read. ## Errors Stay Read-Only If a target is invalid, correct it and retry the same command: ```bash openmeld space board --profile ``` Do not recover an invalid read target by suggesting or running `space create`, `space join`, or `space watch`. They do not repair the malformed input, and the first two can change user state. Preserve Gateway/Auth error codes and request IDs when present. Do not turn an authorization failure, unavailable service, or malformed response into empty data or a successful read. ## Performance And Freshness These commands must read current authorized projections. `space access` is the intentional public preflight and may run without sign-in so Core can return the safe anonymous outcome; content reads still require their normal authorization. Never make reads look fast by caching stale Space data or turning a timeout into success. The CLI may reuse the same already-remotely- validated session within one command while still checking that the locally stored credential is unchanged. If a read is unexpectedly slow, record the exact command, wall time, output mode, and non-sensitive request/error evidence. Retry only the same read; do not write Space state as a diagnostic shortcut. Wake progress is a snapshot for a concrete coordination decision. Do not poll it continuously or infer a terminal outcome when Core still reports active work.