ancoder-skill-cli 0.13.38-beta.1 → 0.13.38-beta.12

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -174,11 +174,24 @@ go build -o bin/skill-cli .
174
174
  | `skill-cli create <name> [--path <dir>]` | Create a skill scaffold with contract and smoke eval templates |
175
175
  | `skill-cli test <path>` | Check that a skill has trigger docs, contract, and eval coverage |
176
176
  | `skill-cli verify <path> [--suite smoke]` | Run a machine-readable verification suite end-to-end |
177
- | `skill-cli generate <name> --desc "..."` | Generate with OMC, then run isolated generator/evaluator contract negotiation and review (default) |
178
- | `skill-cli generate <name> --desc "..." --no-adversarial` | Generate with OMC only, skipping the default adversarial review |
177
+ | `skill-cli generate <name> --desc "..."` | Generate with the default OMC pipeline, real-task feedback, and skill self-loop scaffolding |
178
+ | `skill-cli improve <path-or-name> --logs <dir>` | Improve an existing skill through the default self-iteration loop |
179
+ | `skill-cli loop skill run <path-or-name> --logs <dir>` | Run explicit skill self-iteration cycles with state, budget, and run logs |
180
+ | `skill-cli generate <name> --desc "..." --adversarial` | Also run isolated generator/evaluator contract negotiation and review |
179
181
  | `skill-cli install [--no-omc]` | Install ECC components into `~/.claude/` (includes OMC by default) |
180
182
  | `skill-cli install --component omc` | Install only the bundled OMC multi-agent orchestration layer |
181
183
 
184
+ ## Skill Self-Iteration Loops
185
+
186
+ Generated and improved skills now use a loop-engineering operating layer by default. Each skill can carry:
187
+
188
+ - `LOOP.md` for cadence, gates, and escalation rules
189
+ - `STATE.md` for durable memory across runs
190
+ - `loop-budget.md` for token budget and kill switch
191
+ - `loop-run-log.md` for append-only cycle history
192
+
193
+ Use `skill-cli loop skill audit <skill>` to inspect readiness, `skill-cli loop skill init <skill>` to add the files to an older skill, and `skill-cli loop skill run <skill> --logs ./logs` to run evidence-backed self-improvement cycles.
194
+
182
195
  ## Machine-Readable Skill Layout
183
196
 
184
197
  Task-oriented skills can now include a deterministic verification harness:
@@ -201,10 +214,10 @@ my-skill/
201
214
 
202
215
  ## Adversarial Skill Generation
203
216
 
204
- `skill-cli generate` runs an independent adversarial evaluator pass after the OMC generation pipeline by default:
217
+ `skill-cli generate` uses the OMC pipeline by default. For stricter experimental review, add `--adversarial` to run an independent evaluator pass after the OMC generation pipeline:
205
218
 
206
219
  ```bash
207
- skill-cli generate pdf-to-md --desc "Convert PDF files to Markdown"
220
+ skill-cli generate pdf-to-md --desc "Convert PDF files to Markdown" --adversarial
208
221
  ```
209
222
 
210
223
  This mode uses two isolated Claude CLI contexts:
@@ -214,7 +227,7 @@ This mode uses two isolated Claude CLI contexts:
214
227
 
215
228
  The evaluator fails the run when critical issues are found or the score is below `0.80`. Deterministic gate failures are always converted into critical diff items and cap the score below `0.60`.
216
229
 
217
- Use `--no-adversarial` if you want the older OMC-only generation behavior.
230
+ Use the default command without `--adversarial` for the OMC-only generation path.
218
231
 
219
232
  ### Generation failure logs
220
233
 
@@ -224,7 +237,7 @@ Every `skill-cli generate` run writes a local JSONL session log under:
224
237
  <skills-dir>/.skill-cli/logs/generate/<skill-name>-<timestamp>.jsonl
225
238
  ```
226
239
 
227
- When generation fails, the CLI prints the exact log path. The log records major stages such as Claude binary resolution, scaffold creation, brainstorm probe, loop start/ticks, deterministic gates, repair attempts, and adversarial review failures, including captured command output and error tails for troubleshooting.
240
+ Every generation run prints the exact log path. The log records major stages such as Claude binary resolution, scaffold creation, brainstorm probe, loop start/ticks, deterministic gates, repair attempts, real-task feedback artifacts, and optional adversarial review failures, including captured command output and error tails for troubleshooting.
228
241
 
229
242
  ## Publish to npm
230
243
 
Binary file
Binary file
Binary file
Binary file
Binary file
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "ancoder-skill-cli",
3
- "version": "0.13.38-beta.1",
3
+ "version": "0.13.38-beta.12",
4
4
  "description": "CLI for managing everything-claude-code (ECC) components — agents, skills, commands, rules, hooks, MCP configs. Single binary, all assets embedded.",
5
5
  "bin": {
6
6
  "skill-cli": "bin/skill-cli.js"