fix(agent-advisor): anthropic.model_lineup — new knowledge

[leon1418]

Opened by the knowledge auto-update pipeline. Draft — a human decides.

What changed upstream

We've launched Claude Fable 5.1 (claude-fable-5-1), the successor to Claude Fable 5 for long-running agentic coding, knowledge work, and research, alongside C

Verdict: new_knowledge on anthropic.model_lineup

• was — No entry for Claude Fable 5/5.1 or Mythos 5.1 pricing/specs in the reviewed files; ai-anthropic-to-bedrock.md only mentions "Claude Fable 5" once as an explicit non-default (frontier/Mythos-class pricing, opt-in only), with no ID, price, or context window given.
• now — Claude Fable 5.1 (claude-fable-5-1) and Claude Mythos 5.1 (claude-mythos-5-1) launched: 1M token context (default), 128k max output tokens, always-on adaptive thinking, $10/$50 per MTok (same as Fable 5), cache reads $0.25/MTok. Fable 5.1 available via Claude API, Bedrock, Claude Platform on AWS, Vertex AI, and Microsoft Foundry. Mythos 5.1 is Project Glasswing participants only.

Still true: The recommendation in ai-anthropic-to-bedrock.md to not default to Claude Fable 5 (opt-in only, frontier/Mythos-class pricing) still holds in spirit — Fable 5.1 is likewise a specialized/opt-in-tier model, not a default Bedrock migration target.

This is why the old value is not simply overwritten.

Proposed edits

Each row states its justification. No CI check can catch a badly reworded judgment, so the
"why" column is the only protection a reviewer has — please read it rather than the diff alone.

gcp-to-aws/references/shared/pricing-cache.md:463 — value

- | Claude Fable 5           | 5.00     | 25.00     | 12.50          | 20.00          | 1.00       |
+ | Claude Fable 5.1         | 5.00     | 25.00     | 12.50          | 20.00          | 0.25       |

Why: The announcement states cache reads for Fable 5.1 are cut to $0.25 per MTok (from the prior $1.00 rate implied for Fable 5), so this row's cache-read column must be updated for the new model.

Evidence (verbatim from the announcement): "with cache reads cut to $0.25 per MTok"

gcp-to-aws/references/shared/pricing-cache.md:394 — value

- | Claude Fable 5                   | anthropic.claude-fable-5                 | Anthropic | 10.00      | 50.00       | 1M      | frontier  | active                                                       |
+ | Claude Fable 5.1                 | anthropic.claude-fable-5-1               | Anthropic | 10.00      | 50.00       | 1M      | frontier  | active                                                       |

Why: Fable 5.1 is the successor model now launched with a new model ID (claude-fable-5-1), same $10/$50 pricing and 1M context, so this pricing table row should be updated to reflect the new model/ID as current, per the announcement's explicit launch and pricing details.

Evidence (verbatim from the announcement): "Claude Fable 5.1** (claude-fable-5-1), the successor to Claude Fable 5"

Filter false positives

The announcement scan flagged these files; the judge rejected them:

• agent-advisor/references/decision-refs/model-selection.md

Mirrored to migrate/plugins/migration-to-aws/skills

The same edits were applied to this second copy: 2 applied.

Known limits of this proposal

• Blast radius is not stable between runs. Two runs of the same input returned 9 and 13
  locations and neither was a superset of the other, so this list may be incomplete.
• Paths are relative to advisor/plugins/aws-startup-advisor/skills.
• The pipeline did not touch any file the judge did not name.

[leon1418]

[🤖 AI review 🤖] Reviewed head cda16fa. 2 changed files (+4/−4), 1 commit. Cross-plugin parity PASS — both pricing-cache.md copies are byte-identical (SHA256 3642a7a1d197…). 0 drift allowlist changes.

0 mandatory blockers. 1 Nit.

The Fable 5 → 5.1 successor update is correct: on-demand ($10/$50), cache read ($0.25, down from $1.00), batch ($5/$25), and cache-write columns all match the announcement. Removing Fable 5 rather than keeping both rows is the right call since the announcement identifies 5.1 as a direct successor.

Nit: The narrative header on L386 still references "Claude Fable 5" twice ("Do not default to Claude Fable 5 (frontier)" and "Claude Fable 5 is the most expensive Anthropic model at $10/$50"). Now that the table rows say "Fable 5.1", these two mentions are stale. Low impact — the recommendation is still directionally correct — but worth updating for consistency in a follow-up or this PR.

(Pre-existing, not introduced by this PR: Sonnet 5 status text still says "intro $2/$10 thru Aug 31, 2026" — that intro period expired Sep 1.)

CI: 8/8 SUCCESS (build, gitleaks, bandit, semgrep, checkov). 0 approvals. REVIEW_REQUIRED. Branch: MERGEABLE. MERGE-READY pending approval.

[leon1418]

[🤖 AI review 🤖] Nit: The table row is now "Fable 5.1" but the narrative header above (L386) still says "Do not default to Claude Fable 5 (frontier)" and "Claude Fable 5 is the most expensive Anthropic model." This creates a name inconsistency within the same section — a reader might wonder whether Fable 5 and 5.1 are separate models or the same. Worth updating the two header references to "Fable 5.1" for consistency, either here or in a follow-up.

[jkzietz]

Heads up: the aws-startup-advisor plugin has a new home

The plugin now lives in the Agent Toolkit for AWS, merged to main earlier today:
https://github.com/aws/agent-toolkit-for-aws/tree/main/plugins/aws-startup-advisor

We are beginning plans to deprecate this repository, so that's the copy to build on going forward — changes landed here won't reach customers once distribution repoints. Please re-open this PR against aws/agent-toolkit-for-aws.

Porting your diff: paths move from advisor/plugins/aws-startup-advisor/… to plugins/aws-startup-advisor/…. Two differences to expect:

• The plugin declares a single MCP server there — aws-mcp, the unified AWS MCP Server — and pricing is cache-only, with no live lookup.
• The destination enforces markdownlint's default rule set, which this repo does not. MD036 (emphasis used as a heading) and MD059 (descriptive link text) are the two that usually need fixing.

Happy to help with the move if anything doesn't map cleanly.