@action-llama/skill 0.23.4 → 0.24.1

package/{content/commands/debug.md → skills/debug/SKILL.md}

@@ -1,3 +1,9 @@
+---
+name: debug
+description: Diagnose why an Action Llama agent is failing and suggest fixes. Use when debugging, troubleshooting, or investigating agent errors.
+argument-hint: "[agent-name]"
+---
+
 # Debug a failing agent
 
 Diagnose why an agent is failing and suggest fixes.
@@ -10,16 +16,19 @@ Diagnose why an agent is failing and suggest fixes.
 
 3. Call `al_agents` with the agent name to read its full config and SKILL.md body.
 
-4. Analyze the logs and agent configuration to identify the root cause. Common issues:
+4. For detailed reference on agent commands, exit codes, and runtime context, read `debugging.md` from the `al` skill.
+
+5. Analyze the logs and agent configuration to identify the root cause. Common issues:
    - Missing or expired credentials
    - Malformed SKILL.md frontmatter
    - Instructions that lead to tool errors
    - Rate limiting or API failures
    - Docker/container issues
+   - Resource lock contention
 
-5. Present a diagnosis with:
+6. Present a diagnosis with:
    - **Root cause**: What's going wrong and why
    - **Evidence**: Relevant log lines
    - **Fix**: Concrete changes to make (SKILL.md edits, credential updates, config changes)
 
-6. Offer to apply the fix directly.
+7. Offer to apply the fix directly.

package/{content/commands/iterate.md → skills/iterate/SKILL.md}

@@ -1,3 +1,9 @@
+---
+name: iterate
+description: Run an Action Llama agent, analyze output, and improve its instructions. Repeats up to 3 cycles. Use when iterating on, improving, or refining an agent.
+argument-hint: "[agent-name]"
+---
+
 # Iterate on an agent
 
 Run an agent, analyze its output, and improve its instructions. Repeats up to 3 cycles or until the agent runs cleanly.

package/skills/new-agent/SKILL.md

@@ -0,0 +1,36 @@
+---
+name: new-agent
+description: Create a new Action Llama agent with SKILL.md and config.toml. Use when creating, adding, or scaffolding a new agent.
+argument-hint: "<agent-name>"
+---
+
+# Create a new agent
+
+Create a new Action Llama agent named `$ARGUMENTS`.
+
+## Steps
+
+1. Ask the user for:
+   - **Trigger type**: `schedule` (cron), `webhook`, or both
+   - If schedule: a cron expression (e.g. `*/5 * * * *`)
+   - If webhook: the provider and event (e.g. `github` / `issues.opened`)
+   - **Credentials** needed (e.g. `github_token`, `slack_bot_token`)
+   - A one-sentence description of what the agent should do
+
+2. Use `al_agents` to list existing agents and avoid name conflicts.
+
+3. Create `agents/$ARGUMENTS/SKILL.md` with:
+   - YAML frontmatter containing portable metadata only: `name`, `description`, optionally `license` and `compatibility`
+   - Do NOT include `name` or `model` in the frontmatter — name is derived from the directory, model is inherited from project config
+   - A markdown body with clear instructions for the agent
+   - For the full SKILL.md specification and available agent commands, read `agent-authoring.md` from the `al` skill
+
+4. Create `agents/$ARGUMENTS/config.toml` with runtime configuration:
+   - `models` — list of named model references from project `config.toml`
+   - `credentials` — list of credential types the agent needs
+   - `schedule` and/or `webhooks` — trigger configuration
+   - For the full config.toml reference, read `agent-authoring.md` from the `al` skill
+
+5. Confirm the agent was created and suggest next steps:
+   - `al doctor` to verify credentials
+   - `/al:run` to test it

package/{content/commands/run.md → skills/run/SKILL.md}

@@ -1,3 +1,9 @@
+---
+name: run
+description: Trigger an Action Llama agent run and report the results. Use when running, triggering, or testing an agent.
+argument-hint: "[agent-name]"
+---
+
 # Run an agent
 
 Trigger an agent run and report the results.

package/{content/commands/status.md → skills/status/SKILL.md}

@@ -1,3 +1,8 @@
+---
+name: status
+description: Show a rich overview of the Action Llama project status including scheduler state and agent details. Use when checking status, health, or state of the project.
+---
+
 # Status overview
 
 Show a rich overview of the Action Llama project status.