Microsoft GH-300 Bundle

"Delegate research to a subagent and collect a bounded report" targets A large task has a separate documentation research component. Isolated subtask context avoids flooding the primary task. That differs from A user expects a subagent to remember all parent-session state automatically; the required decision here is "Supply a precise task description and required references". Context boundaries can omit implicit information from the parent.

The action "Delegate distinct self-contained tasks rather than sharing an unbounded thread" is useful when A maintainer wants two unrelated investigations run independently. Separate context reduces cross-contamination of assumptions. Under the stated condition A user expects a subagent to remember all parent-session state automatically, instead choose "Supply a precise task description and required references". Context boundaries can omit implicit information from the parent.

The alternative "Have the main agent reconcile the result against source-of-truth constraints" fits another situation: A subagent suggests a fix contradicting the main application's requirements. Delegated results are recommendations, not automatic authority. Here the condition is A user expects a subagent to remember all parent-session state automatically; use "Supply a precise task description and required references" instead. Context boundaries can omit implicit information from the parent.

The action "Use a sandbox with non-sensitive fixtures and inspect tool results" is useful when A team wants to test an MCP tool before production use. External tools can return malicious or inaccurate instructions. Under the stated condition An IDE agent needs to inspect a supported external issue tracker, instead choose "Configure an authorized MCP integration for that service". MCP exposes tool capabilities beyond local chat context.

The alternative "Validate server compatibility and tool parameter contracts" fits another situation: A feature stops working after an MCP server schema update. Tool shapes must match the current integration API. Here the condition is An IDE agent needs to inspect a supported external issue tracker; use "Configure an authorized MCP integration for that service" instead. MCP exposes tool capabilities beyond local chat context.

The alternative "Limit its credentials and granted tool permissions" addresses An MCP server asks for full repository secrets although only issue titles are needed. Tool integration should enforce least privilege. It does not resolve An IDE agent needs to inspect a supported external issue tracker; the correct choice is "Configure an authorized MCP integration for that service". MCP exposes tool capabilities beyond local chat context.

The alternative "Treat retrieved text as data, not privileged commands" fits another situation: An assistant treats text retrieved through an MCP tool as an instruction to delete files. External tool results do not override trusted operating rules. Here the condition is An MCP server asks for full repository secrets although only issue titles are needed; use "Limit its credentials and granted tool permissions" instead. Tool integration should enforce least privilege.

For the different scenario An IDE agent needs to inspect a supported external issue tracker, "Configure an authorized MCP integration for that service" could be appropriate. MCP exposes tool capabilities beyond local chat context. The present requirement is An MCP server asks for full repository secrets although only issue titles are needed, which calls for "Limit its credentials and granted tool permissions". Tool integration should enforce least privilege.

"Use a sandbox with non-sensitive fixtures and inspect tool results" targets A team wants to test an MCP tool before production use. External tools can return malicious or inaccurate instructions. That differs from An MCP server asks for full repository secrets although only issue titles are needed; the required decision here is "Limit its credentials and granted tool permissions". Tool integration should enforce least privilege.

The alternative "Limit its credentials and granted tool permissions" addresses An MCP server asks for full repository secrets although only issue titles are needed. Tool integration should enforce least privilege. It does not resolve A team wants to test an MCP tool before production use; the correct choice is "Use a sandbox with non-sensitive fixtures and inspect tool results". External tools can return malicious or inaccurate instructions.

For the different scenario An assistant treats text retrieved through an MCP tool as an instruction to delete files, "Treat retrieved text as data, not privileged commands" could be appropriate. External tool results do not override trusted operating rules. The present requirement is A team wants to test an MCP tool before production use, which calls for "Use a sandbox with non-sensitive fixtures and inspect tool results". External tools can return malicious or inaccurate instructions.

The action "Validate server compatibility and tool parameter contracts" is useful when A feature stops working after an MCP server schema update. Tool shapes must match the current integration API. Under the stated condition A team wants to test an MCP tool before production use, instead choose "Use a sandbox with non-sensitive fixtures and inspect tool results". External tools can return malicious or inaccurate instructions.

For the different scenario A team wants to test an MCP tool before production use, "Use a sandbox with non-sensitive fixtures and inspect tool results" could be appropriate. External tools can return malicious or inaccurate instructions. The present requirement is An assistant treats text retrieved through an MCP tool as an instruction to delete files, which calls for "Treat retrieved text as data, not privileged commands". External tool results do not override trusted operating rules.

The action "Validate server compatibility and tool parameter contracts" is useful when A feature stops working after an MCP server schema update. Tool shapes must match the current integration API. Under the stated condition An assistant treats text retrieved through an MCP tool as an instruction to delete files, instead choose "Treat retrieved text as data, not privileged commands". External tool results do not override trusted operating rules.

The alternative "Configure an authorized MCP integration for that service" fits another situation: An IDE agent needs to inspect a supported external issue tracker. MCP exposes tool capabilities beyond local chat context. Here the condition is An assistant treats text retrieved through an MCP tool as an instruction to delete files; use "Treat retrieved text as data, not privileged commands" instead. External tool results do not override trusted operating rules.

"Treat retrieved text as data, not privileged commands" targets An assistant treats text retrieved through an MCP tool as an instruction to delete files. External tool results do not override trusted operating rules. That differs from A feature stops working after an MCP server schema update; the required decision here is "Validate server compatibility and tool parameter contracts". Tool shapes must match the current integration API.

The action "Configure an authorized MCP integration for that service" is useful when An IDE agent needs to inspect a supported external issue tracker. MCP exposes tool capabilities beyond local chat context. Under the stated condition A feature stops working after an MCP server schema update, instead choose "Validate server compatibility and tool parameter contracts". Tool shapes must match the current integration API.

The alternative "Limit its credentials and granted tool permissions" addresses An MCP server asks for full repository secrets although only issue titles are needed. Tool integration should enforce least privilege. It does not resolve A feature stops working after an MCP server schema update; the correct choice is "Validate server compatibility and tool parameter contracts". Tool shapes must match the current integration API.

The alternative "Add path-specific instructions for the Python subtree" fits another situation: Python changes require lint rules different from the repository's TypeScript files. Scoped rules reduce irrelevant instructions on other files. Here the condition is A team wants standing coding standards applied across Copilot interactions; use "Add supported repository-wide Copilot instructions" instead. Persistent instructions provide shared conventions without repeating each prompt.

The alternative "Create a reusable prompt file rather than permanent global guidance" addresses A team wants a focused migration task reusable on demand. Prompt files are invoked for specific tasks rather than automatically everywhere. It does not resolve A team wants standing coding standards applied across Copilot interactions; the correct choice is "Add supported repository-wide Copilot instructions". Persistent instructions provide shared conventions without repeating each prompt.

For the different scenario A repository wants conventions shared with other coding agents, "Use appropriate AGENTS.md guidance alongside Copilot-specific settings" could be appropriate. Agent-wide rules and Copilot instructions serve related but distinct consumers. The present requirement is A team wants standing coding standards applied across Copilot interactions, which calls for "Add supported repository-wide Copilot instructions". Persistent instructions provide shared conventions without repeating each prompt.

The alternative "Use appropriate AGENTS.md guidance alongside Copilot-specific settings" addresses A repository wants conventions shared with other coding agents. Agent-wide rules and Copilot instructions serve related but distinct consumers. It does not resolve Python changes require lint rules different from the repository's TypeScript files; the correct choice is "Add path-specific instructions for the Python subtree". Scoped rules reduce irrelevant instructions on other files.

"Check supported path, branch, and review configuration" targets Copilot code review ignores a newly written instructions file. Instruction discovery is sensitive to file placement and review context. That differs from Python changes require lint rules different from the repository's TypeScript files; the required decision here is "Add path-specific instructions for the Python subtree". Scoped rules reduce irrelevant instructions on other files.

The action "Create a reusable prompt file rather than permanent global guidance" is useful when A team wants a focused migration task reusable on demand. Prompt files are invoked for specific tasks rather than automatically everywhere. Under the stated condition Python changes require lint rules different from the repository's TypeScript files, instead choose "Add path-specific instructions for the Python subtree". Scoped rules reduce irrelevant instructions on other files.

For the different scenario Copilot code review ignores a newly written instructions file, "Check supported path, branch, and review configuration" could be appropriate. Instruction discovery is sensitive to file placement and review context. The present requirement is A team wants a focused migration task reusable on demand, which calls for "Create a reusable prompt file rather than permanent global guidance". Prompt files are invoked for specific tasks rather than automatically everywhere.

"Add supported repository-wide Copilot instructions" targets A team wants standing coding standards applied across Copilot interactions. Persistent instructions provide shared conventions without repeating each prompt. That differs from A team wants a focused migration task reusable on demand; the required decision here is "Create a reusable prompt file rather than permanent global guidance". Prompt files are invoked for specific tasks rather than automatically everywhere.

The action "Use appropriate AGENTS.md guidance alongside Copilot-specific settings" is useful when A repository wants conventions shared with other coding agents. Agent-wide rules and Copilot instructions serve related but distinct consumers. Under the stated condition A team wants a focused migration task reusable on demand, instead choose "Create a reusable prompt file rather than permanent global guidance". Prompt files are invoked for specific tasks rather than automatically everywhere.

"Add supported repository-wide Copilot instructions" targets A team wants standing coding standards applied across Copilot interactions. Persistent instructions provide shared conventions without repeating each prompt. That differs from A repository wants conventions shared with other coding agents; the required decision here is "Use appropriate AGENTS.md guidance alongside Copilot-specific settings". Agent-wide rules and Copilot instructions serve related but distinct consumers.

The action "Add path-specific instructions for the Python subtree" is useful when Python changes require lint rules different from the repository's TypeScript files. Scoped rules reduce irrelevant instructions on other files. Under the stated condition A repository wants conventions shared with other coding agents, instead choose "Use appropriate AGENTS.md guidance alongside Copilot-specific settings". Agent-wide rules and Copilot instructions serve related but distinct consumers.

The alternative "Check supported path, branch, and review configuration" addresses Copilot code review ignores a newly written instructions file. Instruction discovery is sensitive to file placement and review context. It does not resolve A repository wants conventions shared with other coding agents; the correct choice is "Use appropriate AGENTS.md guidance alongside Copilot-specific settings". Agent-wide rules and Copilot instructions serve related but distinct consumers.

The action "Add path-specific instructions for the Python subtree" is useful when Python changes require lint rules different from the repository's TypeScript files. Scoped rules reduce irrelevant instructions on other files. Under the stated condition Copilot code review ignores a newly written instructions file, instead choose "Check supported path, branch, and review configuration". Instruction discovery is sensitive to file placement and review context.

The alternative "Create a reusable prompt file rather than permanent global guidance" fits another situation: A team wants a focused migration task reusable on demand. Prompt files are invoked for specific tasks rather than automatically everywhere. Here the condition is Copilot code review ignores a newly written instructions file; use "Check supported path, branch, and review configuration" instead. Instruction discovery is sensitive to file placement and review context.

The alternative "Add supported repository-wide Copilot instructions" addresses A team wants standing coding standards applied across Copilot interactions. Persistent instructions provide shared conventions without repeating each prompt. It does not resolve Copilot code review ignores a newly written instructions file; the correct choice is "Check supported path, branch, and review configuration". Instruction discovery is sensitive to file placement and review context.