From f62d539f1d9ad33a79275b113b495146a0695748 Mon Sep 17 00:00:00 2001 From: "anthropic-code-agent[bot]" <242468646+Claude@users.noreply.github.com> Date: Thu, 30 Apr 2026 08:52:13 +0000 Subject: [PATCH 1/5] Initial plan From 5e5bb5287bf969b5d6df81ddd8217dc2c8b056c3 Mon Sep 17 00:00:00 2001 From: "anthropic-code-agent[bot]" <242468646+Claude@users.noreply.github.com> Date: Thu, 30 Apr 2026 08:54:00 +0000 Subject: [PATCH 2/5] docs: align agi.prompt.md descriptions with hidden chain-of-thought behavior Agent-Logs-Url: https://github.com/Bryan-Roe/Aria/sessions/7c58f8b9-3482-4a89-9739-fb771082a16c Co-authored-by: Bryan-Roe <74067792+Bryan-Roe@users.noreply.github.com> --- .github/COPILOT_SETUP_GUIDE.md | 4 ++-- .github/copilot-instructions.md | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/.github/COPILOT_SETUP_GUIDE.md b/.github/COPILOT_SETUP_GUIDE.md index 26f023a2e..162aa3555 100644 --- a/.github/COPILOT_SETUP_GUIDE.md +++ b/.github/COPILOT_SETUP_GUIDE.md @@ -382,7 +382,7 @@ Aria includes **20+ specialized prompts** in `.github/prompts/` that guide Copil | Prompt | Purpose | |--------|---------| -| `agi.prompt.md` | Chain-of-thought AGI reasoning | +| `agi.prompt.md` | Autonomous long-running work with internal reasoning | | `debug.prompt.md` | Systematic debugging protocol | | `reason.prompt.md` | Structured analysis & planning | | `review.prompt.md` | Security, performance, correctness review | @@ -394,7 +394,7 @@ Aria includes **20+ specialized prompts** in `.github/prompts/` that guide Copil **Explicit reference**: ``` -Use the agi.prompt for chain-of-thought reasoning about this architecture +Use the agi.prompt for autonomous long-running work requiring internal reasoning Use the debug.prompt to systematically diagnose this error ``` diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md index 2276ec1cd..a016a1d46 100644 --- a/.github/copilot-instructions.md +++ b/.github/copilot-instructions.md @@ -308,7 +308,7 @@ Available agents in `.github/agents/`: |-------|---------| | `ai.agent.md` | Primary autonomous agent — task decomposition, multi-step execution | | `my-agent.agent.md` | QAI specialist — quantum-AI/ML development | -| `agi-reasoning.agent.md` | Chain-of-thought reasoning, self-reflection | +| `agi-reasoning.agent.md` | Structured reasoning with internal chain-of-thought, self-reflection | | `aria-character.agent.md` | Interactive character commands, animations | | `autonomous-trainer.agent.md` | LoRA training lifecycle, model promotion | | `full-stack-debugger.agent.md` | Cross-stack issue diagnosis | @@ -330,7 +330,7 @@ Available agents in `.github/agents/`: - `Azure_Static_Web_App.agent.md` — Static web app deployment patterns **Prompts** (`.github/prompts/`): -- `agi.prompt.md` — AGI reasoning with chain-of-thought +- `agi.prompt.md` — Autonomous long-running work with internal reasoning (hidden chain-of-thought) - `reason.prompt.md` — Structured analysis - `debug.prompt.md` — Systematic diagnostic protocol - `review.prompt.md` — Code review (correctness, security, performance) From 6b94199f7fff4553171408ec4b216bf1d63d4062 Mon Sep 17 00:00:00 2001 From: Bryan <74067792+Bryan-Roe@users.noreply.github.com> Date: Tue, 5 May 2026 21:47:12 -0400 Subject: [PATCH 3/5] Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> Signed-off-by: Bryan <74067792+Bryan-Roe@users.noreply.github.com> --- .github/copilot-instructions.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md index a016a1d46..a1b01ba01 100644 --- a/.github/copilot-instructions.md +++ b/.github/copilot-instructions.md @@ -308,7 +308,7 @@ Available agents in `.github/agents/`: |-------|---------| | `ai.agent.md` | Primary autonomous agent — task decomposition, multi-step execution | | `my-agent.agent.md` | QAI specialist — quantum-AI/ML development | -| `agi-reasoning.agent.md` | Structured reasoning with internal chain-of-thought, self-reflection | +| `agi-reasoning.agent.md` | Structured chain-of-thought reasoning, self-reflection | | `aria-character.agent.md` | Interactive character commands, animations | | `autonomous-trainer.agent.md` | LoRA training lifecycle, model promotion | | `full-stack-debugger.agent.md` | Cross-stack issue diagnosis | From 426aa0e70990b3355b73bfa7c34ca04400889b0a Mon Sep 17 00:00:00 2001 From: Bryan <74067792+Bryan-Roe@users.noreply.github.com> Date: Tue, 5 May 2026 21:47:24 -0400 Subject: [PATCH 4/5] Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> Signed-off-by: Bryan <74067792+Bryan-Roe@users.noreply.github.com> --- .github/copilot-instructions.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md index a1b01ba01..e9c066b80 100644 --- a/.github/copilot-instructions.md +++ b/.github/copilot-instructions.md @@ -330,7 +330,7 @@ Available agents in `.github/agents/`: - `Azure_Static_Web_App.agent.md` — Static web app deployment patterns **Prompts** (`.github/prompts/`): -- `agi.prompt.md` — Autonomous long-running work with internal reasoning (hidden chain-of-thought) +- `agi.prompt.md` — Autonomous long-running work with clear reasoning chains - `reason.prompt.md` — Structured analysis - `debug.prompt.md` — Systematic diagnostic protocol - `review.prompt.md` — Code review (correctness, security, performance) From de9f43c6ba345ed917b5f1685b60476d03ffbe29 Mon Sep 17 00:00:00 2001 From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com> Date: Wed, 6 May 2026 01:50:04 +0000 Subject: [PATCH 5/5] fix: update agi.prompt.md and agi-reasoning.agent.md to enforce internal chain-of-thought Agent-Logs-Url: https://github.com/Bryan-Roe/Aria/sessions/382dfb64-2813-4424-9bb2-b80df22b31ac Co-authored-by: Bryan-Roe <74067792+Bryan-Roe@users.noreply.github.com> --- .github/agents/agi-reasoning.agent.md | 4 +++- .github/copilot-instructions.md | 4 ++-- .github/prompts/agi.prompt.md | 6 ++++-- 3 files changed, 9 insertions(+), 5 deletions(-) diff --git a/.github/agents/agi-reasoning.agent.md b/.github/agents/agi-reasoning.agent.md index 1e636ae6b..02811e774 100644 --- a/.github/agents/agi-reasoning.agent.md +++ b/.github/agents/agi-reasoning.agent.md @@ -1,6 +1,6 @@ --- name: agi-reasoning -description: "AGI reasoning and autonomous decision-making agent. Specializes in chain-of-thought reasoning, task decomposition, self-reflection, and multi-step planning using the AGI provider system.\n\nTrigger phrases include:\n- 'reason through this problem'\n- 'break this down step by step'\n- 'think through this autonomously'\n- 'use AGI reasoning'\n- 'chain of thought'\n- 'self-reflection'\n- 'autonomous planning'\n\nExamples:\n- User says 'reason through this architecture decision' → invoke for structured multi-step analysis\n- User asks 'break down this complex feature into tasks' → invoke for task decomposition\n- User says 'autonomously plan and implement this feature' → invoke for planning + execution with self-correction\n\nThis agent leverages the AGI provider's reasoning chains, task decomposition, and self-reflection capabilities." +description: "AGI reasoning and autonomous decision-making agent. Specializes in internal chain-of-thought reasoning, task decomposition, self-reflection, and multi-step planning using the AGI provider system. Reasoning steps are performed internally and are NOT exposed to the user — only the final answer is returned. For visible step-by-step reasoning, use the visible-reasoning agent instead.\n\nTrigger phrases include:\n- 'reason through this problem'\n- 'break this down step by step'\n- 'think through this autonomously'\n- 'use AGI reasoning'\n- 'chain of thought'\n- 'self-reflection'\n- 'autonomous planning'\n\nExamples:\n- User says 'reason through this architecture decision' → invoke for structured multi-step analysis\n- User asks 'break down this complex feature into tasks' → invoke for task decomposition\n- User says 'autonomously plan and implement this feature' → invoke for planning + execution with self-correction\n\nThis agent leverages the AGI provider's reasoning chains, task decomposition, and self-reflection capabilities. Chain-of-thought is internal only — contrast with visible-reasoning agent which exposes reasoning steps." tools: - edit - search @@ -28,6 +28,8 @@ tools: You are an advanced autonomous reasoning agent built on Aria's AGI provider system. You specialize in structured thinking, task decomposition, self-reflection, and iterative self-improvement. +**Do not expose your internal chain-of-thought in responses.** All reasoning steps (analysis, decomposition, planning, reflection) happen internally. Return only the final answer, decision, or output to the user. For visible step-by-step reasoning, use the `visible-reasoning` agent instead. + ## Return-to-Agent Contract This specialist mode is temporary. After completing the reasoning-heavy portion of the task, return a concise handoff to the primary `agent` that includes: diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md index e9c066b80..ad5a843e0 100644 --- a/.github/copilot-instructions.md +++ b/.github/copilot-instructions.md @@ -308,7 +308,7 @@ Available agents in `.github/agents/`: |-------|---------| | `ai.agent.md` | Primary autonomous agent — task decomposition, multi-step execution | | `my-agent.agent.md` | QAI specialist — quantum-AI/ML development | -| `agi-reasoning.agent.md` | Structured chain-of-thought reasoning, self-reflection | +| `agi-reasoning.agent.md` | Autonomous reasoning with internal chain-of-thought (not exposed); returns final answer only | | `aria-character.agent.md` | Interactive character commands, animations | | `autonomous-trainer.agent.md` | LoRA training lifecycle, model promotion | | `full-stack-debugger.agent.md` | Cross-stack issue diagnosis | @@ -330,7 +330,7 @@ Available agents in `.github/agents/`: - `Azure_Static_Web_App.agent.md` — Static web app deployment patterns **Prompts** (`.github/prompts/`): -- `agi.prompt.md` — Autonomous long-running work with clear reasoning chains +- `agi.prompt.md` — Autonomous long-running work with internal reasoning (chain-of-thought hidden; final answer only) - `reason.prompt.md` — Structured analysis - `debug.prompt.md` — Systematic diagnostic protocol - `review.prompt.md` — Code review (correctness, security, performance) diff --git a/.github/prompts/agi.prompt.md b/.github/prompts/agi.prompt.md index 019f8845d..25cc1715d 100644 --- a/.github/prompts/agi.prompt.md +++ b/.github/prompts/agi.prompt.md @@ -1,11 +1,13 @@ --- -description: "Engage autonomous AGI reasoning with multi-step analysis, task decomposition, self-correction, and iterative improvement. For complex problems requiring structured autonomous thinking." +description: "Engage autonomous AGI reasoning with multi-step analysis, task decomposition, self-correction, and iterative improvement. Reasoning is performed internally — only the final answer is surfaced. For visible step-by-step reasoning, use reason.prompt instead." name: "AGI Reasoning" argument-hint: "Problem or task description (example: analyze the auth flow for race conditions + constraints)" agent: agi-reasoning --- You are an autonomous AGI agent capable of independent reasoning, self-correction, and iterative improvement using the Aria platform's AGI provider system. +**Do not expose your internal chain-of-thought in responses.** Reasoning happens internally — deliver only the final answer, decision, or output. + **Reasoning Pipeline:** 1. **Analyze** — Classify the task: @@ -47,5 +49,5 @@ create_agi_provider(reasoning_depth=3, enable_chain_of_thought=True, enable_self **Success Criteria:** - Solution is correct, complete, and verified - Follows existing codebase patterns and conventions -- Clear reasoning chain showing decision process +- Reasoning is performed internally; only the final answer/output is returned - Minimal change surface (no unnecessary modifications)