From 051bac4183ac86c3cf647c660caecd0cb67fa781 Mon Sep 17 00:00:00 2001 From: Levi Neely Date: Thu, 16 Apr 2026 21:18:18 +0200 Subject: [PATCH] execute: update skills for execute_tool removal Co-Authored-By: Claude Sonnet 4.6 --- execute-pipe/SKILL.md | 7 +++---- pcloudcc-testing/SKILL.md | 21 +++++---------------- 2 files changed, 8 insertions(+), 20 deletions(-) diff --git a/execute-pipe/SKILL.md b/execute-pipe/SKILL.md index 78f0ca8..e7ff0a7 100644 --- a/execute-pipe/SKILL.md +++ b/execute-pipe/SKILL.md @@ -1,7 +1,7 @@ --- name: execute-pipe intent: parallel, pipeline, pipe, fan-out, concurrent, batch, etl, transform, compose -description: Compose execute_code and execute_tool into sequential pipelines with optional parallel fan-out stages. Use for multi-step data processing, parallel independent operations, and ETL workflows. +description: Compose execute_code into sequential pipelines with optional parallel fan-out stages using execute_pipe. Use for multi-step data processing, parallel independent operations, and ETL workflows. --- # Execute Pipeline Patterns @@ -14,9 +14,8 @@ The primitives stay simple. Complexity lives in the data schema and the skill (h | Tool | What it does | |---|---| -| `execute_code` | Run one or more code snippets. Multiple steps run in parallel; single step returns raw output. | +| `execute_code` | Run one or more steps. Steps may be inline `{code}` or named `{tool, args}` scripts. Multiple steps run in parallel; single step returns raw output. | | `execute_pipe` | Chain stages sequentially, piping stdout → stdin. Each stage: `{code}`, `{tool, args}`, or `{parallel: [steps]}`. | -| `execute_tool` | Run a named script from `ollie/t`. Used inside `execute_pipe` as a `{tool}` stage. | ## execute_code: parallel fan-out @@ -120,4 +119,4 @@ If normalization requires multiple steps, extract it as a named tool in `ollie/t | N independent operations, same output schema | `execute_code` parallel steps, or `execute_pipe` parallel stage | | N operations feeding one transform | `execute_pipe` with parallel first stage | | Multi-step data transform | `execute_pipe` sequential stages | -| Reusable transform logic | `execute_tool` or named script in `ollie/t` | +| Reusable transform logic | `{tool}` step in `execute_code`, or named script in `ollie/t` as a pipe stage | diff --git a/pcloudcc-testing/SKILL.md b/pcloudcc-testing/SKILL.md index b5874da..fb17278 100644 --- a/pcloudcc-testing/SKILL.md +++ b/pcloudcc-testing/SKILL.md @@ -70,31 +70,22 @@ Load pcloudcc-testing skill. Test the affected code path with unit tests, fault Spawn testing agent: ``` -Tool: execute_tool -tool: spawn_agent.sh -args: ["", "", ""] -sandbox: default +execute_code: steps=[{tool: "spawn_agent.sh", args: ["", "", ""]}] ``` Send test request: ``` -Tool: execute_tool -tool: send_message.sh -args: ["", "", "PROMPT_REQUEST", "Validate fix", "Bug: \nFix: \nFiles: \nBranch: "] +execute_code: steps=[{tool: "send_message.sh", args: ["", "", "PROMPT_REQUEST", "Validate fix", "Bug: \nFix: \nFiles: \nBranch: "]}] ``` Read response (user notifies when ready): ``` -Tool: execute_tool -tool: check_inbox.sh -args: [""] +execute_code: steps=[{tool: "check_inbox.sh", args: [""]}] ``` Kill testing agent: ``` -Tool: execute_tool -tool: kill_agent.sh -args: [""] +execute_code: steps=[{tool: "kill_agent.sh", args: [""]}] ``` ### Testing Agent Workflow @@ -106,9 +97,7 @@ args: [""] 5. Verify fix (old fails, new passes) 6. Report: ``` -Tool: execute_tool -tool: send_message.sh -args: ["", "", "PROMPT_RESPONSE", "Validation complete", "Status: completed|failed\n"] +execute_code: steps=[{tool: "send_message.sh", args: ["", "", "PROMPT_RESPONSE", "Validation complete", "Status: completed|failed\n"]}] ``` **FUSE testing**: Use `execute_elevated_bash` to start the daemon outside the sandbox (see Sandbox Escape section above). No user intervention required.