Repository navigation
ASE 2.0.7
2.0.7 (2026-11-04)
-
FEATURE [code, docs]: Model Selection of Agents
The new definition option-M|--modelof all agents (alsomodelof the definition files and
of the sectionsdefaultand per agent of a fleet) selects the model of the agent tool, either
for all tasks, or as<size>=<model>pairs perEffort:size of the task (like
XS=haiku,*=sonnet), while the new hookmodelof a definition can override it per task. -
FEATURE [code, docs]: Partial Staging Skill
The new skillase-repo-stagestages only the working copy changes belonging to a topic, the
lastCHANGELOG.mdentry (--entry), the last turn (--turn), or the session (--session),
splitting mixed hunks, checks that the staged state builds on its own (optionally in a temporary
worktree), and optionally chains intoase-repo-changelogandase-repo-commit. -
IMPROVEMENT [code, docs]: Verification Levels for Edit Skills
The boolean option--verify|-vofase-code-editandase-spec-editnow takes a level:
none(former default),light(new default: lint, build, and regression tests resp. SpecBook
lint), orfull(additionally ad-hoc manual tests), also steering the grilling questions. Both
skills now fix and re-verify for at most ten rounds and list the remaining failures. -
IMPROVEMENT [code, docs]: ChangeLog Entries in File Style
The skillase-repo-changelognow follows the style of the existing entries instead of a fixed
60-80 character summary, keeps each entry within about 5 lines, writes the section viaEdit
(so its changes show up as a diff), and triggers on prompts like "add a changelog entry". -
IMPROVEMENT [code]: Free-Text Replies in Next-Step Dialogs
The next-step dialogs ofase-task-implement,ase-task-integrate,ase-task-grill,
ase-task-reboot, andase-task-condensenow accept a free-text reply as an implicitDONE:
the skill finishes, and the text is then handled as a regular request, instead of the dialog
being re-opened with an "Invalid option selected!" error. -
IMPROVEMENT [code]: Commit Message Skill Activation
The frontmatter description ofase-repo-commitnow lists trigger phrases (like "write" or
"propose" a "commit message"), so the skill is activated under more circumstances. -
IMPROVEMENT [code]: find(1) Fallback for Source Resolution
ase-code-lintnow resolves directory and wildcard source references withfindviaBash
(newly allowed) if theGlobtool is not available, to support foreign LLMs. -
BUGFIX [code]: Gitea Issue Dependencies Content-Type
The Gitea task store plugin (ase-task-store-plugin-gitea.ts) now sends the issue dependency
create and remove requests withContent-Type: application/json, fixing the HTTP 422 errors
against Forgejo/Gitea. -
BUGFIX [code]: Silent parallel sub-agents
Forbid any output while concurrently dispatched sub-agents report back one by one, as Claude
Code runs them in the background despiterun_in_background: false, so skills like
ase-code-lintandase-docs-proofreadno longer print progress messages per returned batch
(ase-control.md). -
BUGFIX [code]: Skill Started Message First
ase-skill.mdnow requires the "skill started" message as the very first output, before any
tool call (includingase_getopt) and also for skills invoked via theSkilltool, so that
"do not output anything" instructions of later steps no longer suppress or defer it. -
BUGFIX [code]: Synopsis Band Continuation Indentation
The continuation lines of the skill synopsis in the hint band are now indented by 2 instead of
4 spaces, so they align with the command after the prompt marker. -
UPDATE [infr]: NPM dependencies
Updated NPM dependencies.