agent-loop-ext
Crash-resilient external agent loop with state persistence and CI/CD integration
Validate a single AIWG component (skill, agent, or command) for completeness and correctness
$ npx -y skills add jmagly/aiwg --skill validate-component --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/validate-componentContext preview
The summary Claude sees to decide when to auto-load this skill.
Validate a single AIWG component (skill, agent, or command) for completeness and correctness
namespace: aiwg name: validate-component description: Validate a single AIWG component (skill, agent, or command) for completeness and correctness platforms: [all]
You check a single AIWG component — a skill, agent, or CLI command — for completeness and correctness, then produce a structured pass/fail report with specific gaps listed.
| Pattern | Example | Action | |---------|---------|--------| | Validate skill | "validate this skill" | Run skill completeness checks | | Validate agent | "check this agent definition" | Run agent completeness checks | | Validate command | "is this command wired up" | Run command completeness checks | | Path-specific | "validate component at agentic/code/addons/my-addon/skills/my-skill" | Validate at given path | | Pre-PR | "pre-PR check on this component" | Run all applicable checks |
1. **Identify component type and path**:
2. **Run type-appropriate checks**:
**For a Skill**:
**For an Agent**:
**Link Classification** (applies to all component types):
Extract every `@<path>` reference from the file and classify:
| Pattern | Result | Message | |---------|--------|---------| | `@$AIWG_ROOT/<path>` | PASS | AIWG core ref (install-relative) | | `@$TOKEN/<path>` — TOKEN set in env | PASS | Registered corpus token | | `@$TOKEN/<path>` — TOKEN NOT in env | WARN | `env var TOKEN not set; add to .env or export` | | `@.aiwg/<path>` — path in Tier 1/2 allowlist | PASS | Normalized project memory | | `@.aiwg/<path>` — not in any `memory.creates` | FAIL | `non-normalized .aiwg/ path — repo-local only` | | `@.claude/<path>` | FAIL | `deployment target — forbidden in distributable source` | | `@agentic/code/<path>` | WARN | `legacy bare ref — migrate to @$AIWG_ROOT/agentic/code/` | | `@src/<path>` | WARN | `legacy bare ref — migrate to @$AIWG_ROOT/src/` | | `@docs/<path>` | WARN | `legacy bare ref — migrate to @$AIWG_ROOT/docs/` | | `@tools/<path>` | WARN | `legacy bare ref — migrate to @$AIWG_ROOT/tools/` | | Relative path within component | PASS | Local ref |
Build the Tier 2 allowlist by reading all `manifest.json` files in `agentic/code/` and collecting `memory.creates[*].path` values. Combine with Tier 1 (`.aiwg/AIWG.md`, `.aiwg/frameworks/`).
**For a CLI Command**:
3. **Compile report**:
Format results as a structured report:
Component Validation: <component-name> (<type>)
Path: <absolute-or-relative-path>
Checks:
PASS description frontmatter present
PASS title section present
PASS behavior section present
FAIL no examples section found
PASS listed in manifest.json
FAIL not in agentic/code/ (found in .claude/skills/ — placement violation)
WARN @.aiwg/planning/my-design.md is not a normalized path (repo-local only)
Result: FAIL — 2 issues found
Required actions:
1. Add ## Examples section to SKILL.md
2. Move SKILL.md to agentic/code/addons/<addon>/skills/<name>/SKILL.md
then re-deploy with: aiwg use <addon>4. **Surface the report** to the user with the result line prominent.
**User**: "Validate the doctor skill"
**Action**: Read `agentic/code/addons/aiwg-utils/skills/doctor/SKILL.md`, find parent manifest, run all checks.
**Output**:
Component Validation: doctor (skill) Path: agentic/code/addons/aiwg-utils/skills/doctor/SKILL.md Checks: PASS description frontmatter present PASS title section present PASS behavior section present PASS examples section present PASS listed in aiwg-utils manifest.json PASS lives in agentic/code/ Result: P
Reusable project context and specialist workflows for the AI tools you already use. Plan software, coordinate specialist reviews, prepare campaigns, investigate incidents, organize research, curate media, and maintain operational knowledge.
Repo: jmagly/aiwg
Crash-resilient external agent loop with state persistence and CI/CD integration
Detect requests for iterative autonomous agent loops and route to the appropriate loop executor
Automatically execute tests when code-generating agents modify source files, enforcing the execute-before-return pattern
Enable agent loops to learn from similar past tasks and share patterns across loops
Query and manage the executable feedback debug memory
Execute tests on generated code and iterate until passing