Fable 5 prompt-layer refresh: fable routing pins + agent reduction - #270
Open
Michael J. Jabbour (michaeljabbour) wants to merge 2 commits into
Open
Fable 5 prompt-layer refresh: fable routing pins + agent reduction#270Michael J. Jabbour (michaeljabbour) wants to merge 2 commits into
Michael J. Jabbour (michaeljabbour) wants to merge 2 commits into
Conversation
Update all agents and common-agent-base.md to optimize for Anthropic's claude-fable-5 model. Anthropic guidance indicates that prompts written for weaker models are often too prescriptive and degrade output on stronger models. This refresh follows that guidance. Changes: - All 15 agents with provider_preferences now list `claude-fable-*` as first preference (existing opus/sonnet wildcards retained as fallbacks); previously no wildcard matched fable-5, mechanically excluding it from agent selection. - modular-builder.md: 611 → 98 lines (removed duplicated templates, checklists, slogan lists, and duplicate @-mention block; agent contract preserved; footer @-mentions trimmed 7 → 3) - zen-architect.md: 424 → 77 lines (removed 3 duplicated section templates, collapsed structure, trimmed footer @-mentions from 7 → 3) - common-agent-base.md: 231 → 196 lines (removed duplicated defensive-security paragraph, collapsed redundant headers, added two fable-5-recommended one-liners on scope discipline and turn-ending clarity) - Repaired corrupted merge splice in modular-builder.md - Net diff: 60 insertions, 927 deletions - Agent-metadata test suite: 24/24 passing (no contracts broken) Generated with [Amplifier](https://github.com/microsoft/amplifier) Co-Authored-By: Amplifier <240397093+microsoft-amplifier@users.noreply.github.com>
…-2,475 lines) Reduced bug-hunter, ecosystem-expert, explorer, file-ops, foundation-expert, git-ops, integration-specialist, post-task-cleanup, security-guardian, session-analyst, shell-exec, test-coverage, web-research from 3,685 to 1,210 total lines (-67%). Removed: rigid output templates, checklists, forbidden-phrase lists, generic code examples, duplicated @-mention blocks/warnings. Preserved: frontmatter, each agent's distinctive methodology, all safety protocols (git-ops destructive-command rules, session-analyst events.jsonl warnings, security-guardian severity taxonomy), repository-conventions-discovery sections, commit/PR format templates. Agent-metadata tests green (24/24) after every file. Generated with [Amplifier](https://github.com/microsoft/amplifier) Co-Authored-By: Amplifier <240397093+microsoft-amplifier@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
Anthropic's claude-fable-5 guidance: prompts written for weaker models are over-prescriptive and degrade output on newer models. Additionally, no existing
provider_preferenceswildcard (e.g.,claude-opus-*/claude-sonnet-*) matchedclaude-fable-5, mechanically excluding it from all 15 agents.Round 1: Prompt-Layer Refresh (commit cac56b1)
claude-fable-*as first provider preference in 15 agents (old pins kept as fallbacks)Round 2: Agent Reduction (commit a425df9)
Verification
Summary
Reviewer callout: One judgment call in common-agent-base.md — of two near-duplicate defensive-security paragraphs, the fuller version (authorized CTF/pentest allowed) was kept over the terser 'defensive only' one. Please flag if the stricter wording was intentional policy.