docs: add background agent files to AGENTS.md; update prompts; add background_agent tool and tests; detached launcher in process-runner; status counts

This commit is contained in:
pi
2026-04-12 14:03:11 +01:00
parent 6c87a9a1a2
commit 4919833b4e
6 changed files with 16 additions and 21 deletions

View File

@@ -1,10 +1,8 @@
---
description: Implement, review, then revise using generic subagents
description: Implement, review, then revise using preset-driven subagents
---
Use the `subagent` tool in chain mode:
1. Run a generic subagent to implement: $@
2. Run a generic subagent to review `{previous}` and identify issues
3. Run a generic subagent to revise the implementation using `{previous}`
Run `subagent` to implement the requested change using an implementation preset.
Then run `subagent` to review the implementation and identify issues.
Finally run `subagent` to revise the implementation using the review output.
User request: $@

View File

@@ -1,10 +1,8 @@
---
description: Inspect, plan, and implement using generic subagents
description: Inspect, plan, and implement using preset-driven subagents
---
Use the `subagent` tool in chain mode:
1. Run a generic subagent to inspect the codebase for: $@
2. Run a generic subagent to turn `{previous}` into a concrete implementation plan for: $@
3. Run a generic subagent to implement the approved plan using `{previous}`
Run `subagent` to inspect and gather context using an appropriate preset.
Then run `subagent` to produce a concrete implementation plan from that context.
Finally run `subagent` to implement the approved plan.
User request: $@

View File

@@ -1,9 +1,7 @@
---
description: Inspect the codebase, then produce a plan using generic subagents
description: Inspect the codebase, then produce a plan using preset-driven subagents
---
Use the `subagent` tool in chain mode:
1. Run a generic subagent to inspect the codebase for: $@
2. Run a generic subagent to turn `{previous}` into a concrete implementation plan for: $@
Run `subagent` to inspect the codebase with an appropriate preset.
Then run `subagent` again with a planning preset, using the first result as context.
User request: $@