Adds the third Managed Agents permission policy, `auto`, alongside
`always_allow` / `always_ask`: the three outcomes (runs, denied as
high-risk with an error tool result while the session keeps running,
pauses for approval when indeterminate), a config example, what the
evaluation trusts, and the "not a human checkpoint" warning. Documents
the `evaluated_permission` and `evaluation` fields on `agent.tool_use` /
`agent.mcp_tool_use`, and updates the client-pattern and multiagent
guides to gate on `evaluated_permission === 'ask'` rather than the
configured policy.
Adds `ant beta:sessions connect` to the CLI guide: terminal viewer
keybindings, the allow/deny prompt, and the `--web` local session viewer.
Fixes the deny example to use `deny_message` (the real field) instead of
`message`.
Claude-Session: https://claude.ai/code/session_01UkZpc4FqLFPLBJF2Zcuq3a
Co-authored-by: Claude <noreply@anthropic.com>
* Update claude-api skill: Managed Agents self-hosted memory stores and web tool domain settings, cost-optimize subcommand, Admin API reference, ASCII-only text
Managed Agents: self-hosted sandboxes can now attach memory stores via the SDK worker (handle_item, ANTHROPIC_WORK_SECRET, sync options, troubleshooting); web_search/web_fetch accept allowed_domains/blocked_domains/user_location/max_content_tokens on the toolset configs entry and are not governed by environment networking; typed per-tool config unions; Console session viewer notes; packages caveat under limited networking; Claude Platform on AWS self-hosted worker auth.
New shared/cost-optimization.md backing a cost-optimize subcommand, and new shared/admin-api.md covering client.beta.organization in all SDKs and the CLI. Prompt caching gains TTL selection, automatic vs explicit breakpoint guidance, workspace isolation, and verification guidance. Sonnet 5 pricing is the permanent $2/$10 list price. Advisor pairing no longer excludes Claude Fable 5 for Managed Agents. Reviewer-only HTML comments are stripped from the published files.
All files are now plain ASCII in prose (em dashes, arrows, emoji callouts, and box-drawing replaced with ASCII equivalents), matching the source so future syncs diff cleanly.
No-Verification-Needed: documentation-only change to skill reference content
* Update claude-api skill: Claude Fable 5.1 / Mythos 5.1 catalog rows, Files and Skills APIs out of beta
Claude Fable 5.1 (claude-fable-5-1) and Claude Mythos 5.1 (claude-mythos-5-1) become the default Fable-tier models throughout the skill; Claude Fable 5 and Mythos 5 stay selectable by id with their own catalog rows. Feature-support lists that named Fable 5 now read Fable 5/5.1.
The Files API and Skills API are out of beta: examples use client.files.* / client.skills.* with no beta header, and the API-drift table points at the beta-to-GA migration docs.
No-Verification-Needed: documentation-only change to skill reference content
* Update claude-api skill: Claude Fable 5.1 / Mythos 5.1 migration section and API changes
Adds a "Migrating to Claude Fable 5.1 from Claude Fable 5" section to shared/model-migration.md: three breaking changes (forced tool_choice any/tool returns 400; thinking blocks are preserved only for the model that produced them or a newer one; and only in the conversation that produced them, so edited history replayed with thinking blocks is rejected), what carries over from Fable 5, the Opus 5 path, Mythos 5.1 notes, capability improvements, prompt-tunable behavioral shifts, and a migration checklist.
New API features documented: per-message effort (mid-conversation-output-config beta), turn-scoped mid-conversation system messages with clear_at, progress updates between tool calls via thinking.display "updates", thinking block_binding controls, and the 0.025x cache-read rate on Fable 5.1 with a max_tokens: 0 keep-alive that usually beats the 1-hour TTL.
Error catalog, prompt-caching, tool-use, platform-availability, cost-optimization, and prompt-audit are updated to match; SKILL.md routes migration and prompting questions to the new section.
No-Verification-Needed: documentation-only change to skill reference content
* Update claude-api skill: Claude Fable 5.1 launch-day hedges and migration-path table rows
Adds claude-fable-5 -> claude-fable-5-1 and claude-mythos-5 -> claude-mythos-5-1 rows to the migration-path and model-ID mapping tables (including the Bedrock IDs), and updates the refusal-fallback example to the 5.1 model id.
Hedges three claims until the launch docs confirm them: Task Budgets support on Claude Fable 5.1, whether Claude Mythos 5.1 shares the 0.025x cache-read rate, and the fallback-credit wording. The block_binding error row now says to send the controls beta header only where that beta is offered and to fall back to strip-and-retry elsewhere. Cross-references within the migration section point at the history-editing check directly.
No-Verification-Needed: documentation-only change to skill reference content
Mirror the latest bundled-skill content for the August 5 Managed Agents
launch wave:
- Session budgets: budget object at session create (minor-unit cents
string), budget_reached pause semantics, settle-event allowlist,
session.usage event, change/remove-to-resume, multiagent shared cap
- Deployment budgets: same budget object on deployments, copied onto
each fired session, clearable and re-addable
- Inference geo pinning: inference_geo on the agent model object,
per-session override, roster uniformity, never grandfathered
- Skills from a GitHub repository: root .claude/skills discovery at
session start (cloud sandboxes only), trust-boundary warning
- Advisor: {type: "advisor", model} multiagent roster entry, reserved
anthropic.advisor thread, plaintext vs redacted delivery; advisor
tool max_uses/max_tokens/caching options and error result variant
- Multiagent: new when-to-use guidance (start with self, add cheaper
workers, dedicated specialists), delegation tools, docs URL rename
- Corrections: agent versions are sequential integers, Files API
uploads take a purpose param, stop_details refusal categories are an
open set, Sonnet 5 added to the prefill-removed list, vault_ids is
create-only on sessions, deployments gain an update endpoint
No-Verification-Needed: markdown-only skill content sync
* Update claude-api skill: Claude Opus 5
- Add Claude Opus 5 (`claude-opus-5`) as the default model across all
SDK examples, model tables, pricing, and migration guidance.
- Add an Opus 5 migration section covering the API changes that come
with it, and move Opus 4.8 into the previous-generation slot.
- Drop the Opus 4.7 fast-mode guidance, which no longer applies.
- Scope the server-side fallbacks beta header to the current version.
* Update claude-api skill: Opus 5 fast-follow corrections
- Server-side refusal fallbacks are available on Claude Platform on AWS,
not just the Claude API. Correct the availability statements in
SKILL.md, the migration guide, and the platform-availability table.
- Simplify the effort guidance: start at `high` (the API default) and
sweep down, rather than starting at `xhigh` for coding work.
- Add a time-to-first-token section with the prompt instruction that
reduces pre-answer thinking on latency-sensitive routes.
- Replace the two separate thinking-disabled mitigations with the single
combined instruction that covers both failure modes.
Syncs the claude-api skill with the current upstream source.
## Managed Agents — five new features
- `effort` on the agent's `model` object (a level string or `{"type": "<level>"}`).
It is agent-configuration only: setting it in a per-session `model` override is
silently ignored.
- Optional `version` on agent update, for optimistic concurrency. Omit it for
last-write-wins.
- `initial_events` on session create, collapsing create plus first send into one
call. Validation is all-or-nothing and only `user.message` and
`user.define_outcome` are accepted.
- Environment and memory-store webhooks: four `environment.*` events and three
`memory_store.*` events.
- Event deltas on per-thread streams. Previews are thread-scoped, so a child
thread's previews never reach the session-level stream.
## Corrections to behavior the skill already documented
- Tool output offload triggers at 100,000 characters (~25k tokens), not 100K
tokens, and covers built-in tools rather than MCP alone.
- `system.message` works on four models, checks only the primary model, appends
system context instead of replacing the prompt, and is accepted during a
`requires_action` idle when it trails a tool result in the same request.
- Vault-to-MCP credential matching is normalized (scheme and host lowercased,
default ports and trailing slashes stripped), not byte-exact.
- Multiagent depth greater than 1 is a validation error, not silently ignored.
- Outcome `interrupted` fires even when evaluation never started, and then
carries an empty-string `outcome_evaluation_start_id`.
- Deployment jitter is 15% of the run interval, floor 5s, cap 9 minutes.
- Webhooks retry three times with jittered 5-120s backoff, then drop silently.
Auto-disable is duration-based with three named triggers.
- `session.status_terminated` means completion or error.
- `agent.thinking` is a progress signal and carries no thinking content.
- `processed_at` is already populated on first sighting for
`user.define_outcome`, `user.custom_tool_result`, and `user.tool_result`.
- Session creation does not provision the sandbox.
- Skill `version` applies to Anthropic-authored skills too.
- Console-created vault credentials are header-injection only, and vault
environment-variable substitution skips secrets in URL paths.
- Console trace URLs need the real workspace ID when the API key is not in the
Default workspace.
## Elsewhere in the skill
- Note partner pricing under the first-party price table: Microsoft Foundry
bills at standard API rates, while Amazon Bedrock and Vertex AI are
partner-operated with separate pricing.
- Correct the tool-runner human-in-the-loop guidance, which previously pointed
readers at the manual loop for approval gates the runner's per-turn hooks
already cover.
- Repoint links away from the retired documentation hosts.
- Dedupe eagerly loaded SKILL.md content that the per-section files already
cover.
