agentpack-cli 0.1.13 → 0.1.15

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/docs/ROADMAP.md CHANGED
@@ -95,58 +95,18 @@ Useful signals:
95
95
 
96
96
  The target is reliable continuation of the work that was worth preserving.
97
97
 
98
- ## Near-Term Roadmap
98
+ ## Development Direction
99
99
 
100
- ### Now: Post-v0.1.12 Dogfood
100
+ Agentpack should grow by proving repo-native continuity in real work before adding orchestration breadth.
101
101
 
102
- Goal: dogfood the released v0.1.12 package and the local post-release Task Passport ergonomics patch before deciding whether to cut v0.1.13.
102
+ Priorities:
103
103
 
104
- Recent dogfood confirmed these workflow pains and fixes:
105
-
106
- - `task audit` was too noisy, so metadata drift is now separated from action-required task warnings
107
- - quick current-task inspection was missing, so `agentpack task status` now exists
108
- - switching chats or agents needed a compact task-first summary, so local main now has `agentpack task handoff`
109
- - ending a task required too many manual steps, so local main now has `agentpack task finalize`
110
- - the release flow is still manual, but it is tolerable enough to keep release automation deferred
111
- - ledger cadence still needs dogfood attention so agents avoid recording every micro-step
112
-
113
- Patch only blockers or cohesive ergonomics improvements during this period. Prefer safe, reviewable chunks that move the product forward without turning every tiny observation into a release.
114
-
115
- ### Current Patch: Task Passport Ergonomics
116
-
117
- Implemented locally after v0.1.12:
118
-
119
- - `agentpack task status`: short human-readable current task view
120
- - `agentpack task handoff`: compact current-passport handoff for chat/client/worktree switches
121
- - `agentpack task finalize`: a small ritual for evidence-backed verification and close
122
- - softer audit output that separates action-required issues from accepted metadata warnings
123
-
124
- Dogfood these together before deciding on a v0.1.13 release. Do not describe `task finalize` as available in the published v0.1.12 package.
125
-
126
- ### Release Flow
127
-
128
- If release friction stays real after another release:
129
-
130
- - add a `release:patch` helper or script
131
- - make it do the version bump, checklist reminder, and maybe local preflight
132
- - do not automate GitHub Actions or npm polling unless explicitly requested
133
-
134
- ### Docs And Product
135
-
136
- After dogfood:
137
-
138
- - update this roadmap to reflect the real post-Task-Passport state
139
- - document the recommended multi-chat workflow: implementation, design review, and teach/learn
140
- - simplify README positioning around the idea that Agentpack is a continuity ledger, not general AI memory
141
-
142
- ### Later
143
-
144
- Only after the short dogfood and ergonomics pass:
145
-
146
- - role lanes inside Task Passport
147
- - better source-cache refresh ergonomics
148
- - structured handoff between separate chats
149
- - maybe `agentpack task handoff` as a compact current-passport-first export
104
+ - make Task Passport handoff reliable and low-friction
105
+ - keep source-cache metadata useful without making audits noisy
106
+ - define workstream separation only after single-passport dogfood is stable
107
+ - support explicit, inspectable handoff/export between chats, clients, worktrees, and machines
108
+ - add orchestrator recipes after the local CLI/MCP contract is boring and stable
109
+ - keep release and ledger discipline simple enough to use during real work
150
110
 
151
111
  ## v0.1: Usable Manual CLI
152
112
 
@@ -185,7 +185,7 @@ agentpack task finalize
185
185
  agentpack task --help
186
186
  ```
187
187
 
188
- `resume` and MCP `load_context` read the current passport automatically when one exists, then show the broader repo-level ledger below it.
188
+ `resume` and MCP `load_context` read the current passport automatically when one exists, then show the broader repo-level ledger below it. If the current pointer still references a closed passport, `resume` labels its next actions as historical so they are not mistaken for active instructions.
189
189
 
190
190
  The normal human-facing sequence is: start the task, keep status/scope/next actions current, record verification with evidence, print a handoff when another agent or chat may continue, then finalize only after verification is final.
191
191
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "agentpack-cli",
3
- "version": "0.1.13",
3
+ "version": "0.1.15",
4
4
  "description": "Local task-state ledger for AI coding agents.",
5
5
  "type": "module",
6
6
  "repository": {