You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: templates/base/docs/live/progress.md
+17-10Lines changed: 17 additions & 10 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -4,34 +4,41 @@ Read after `docs/live/current-focus.md` to recover the latest state, continuity,
4
4
5
5
## Current State
6
6
7
-
Summarize the current state of the work in 1-2 sentences.
7
+
The template skill surface now has a complete top-level discoverability pass: `using-agent-practices` and its category map cover every live top-level skill, including standalone specialist leaves that were previously omitted. No package moves or new router families were needed, so the default next move still returns to Task 35 unless a concrete routing regression appears.
8
8
9
9
## Latest Completed Work
10
10
11
-
- completed item
11
+
- completed a full scan of `templates/base/.agents/skills/` and found six live top-level skills missing from the canonical top-level router surface: `create-skill`, `create-router-skill`, `cx-ticket-triage`, `data-exploration`, `visualization`, and `media`
12
+
- updated `templates/base/.agents/skills/using-agent-practices/{SKILL.md,references/category-map.md,evals/*}` so those skills are first-class direct routes with explicit boundary language against nearby skills
13
+
- updated `templates/base/docs/reference/{codemap.md,memory.md}` to preserve the top-level discoverability rule and the category-map role
14
+
- reviewed reference writeback: no `templates/base/docs/reference/{architecture.md,lessons.md}` update was needed because the change tightened top-level discoverability rules without moving package boundaries or adding a new failure-pattern worth preserving
12
15
13
16
## In Progress
14
17
15
-
- none
18
+
None.
16
19
17
20
## Blockers
18
21
19
-
- none
22
+
None.
20
23
21
24
## Next Recommended Action
22
25
23
-
- Next step:
26
+
- Next step: Resume Task 35: refine the capability-based methodology overlays in the finance, research, and webapp docs. Treat the top-level skill audit as landed unless a concrete discoverability regression appears.
- Check: targeted Python assertions for top-level skill coverage, router outputs, category-map inventory completeness, and eval IDs
39
+
- Result: passed
33
40
34
41
## Hand-off Note
35
42
36
-
- Resume from:
37
-
- Watch for:
43
+
- Resume from:`templates/base/.agents/skills/using-agent-practices/` if any top-level routing ambiguity resurfaces; otherwise return to Task 35 work.
44
+
- Watch for: future standalone skills added under `templates/base/.agents/skills/` without matching updates to `using-agent-practices/SKILL.md`, `references/category-map.md`, and trigger/task eval coverage.
Copy file name to clipboardExpand all lines: templates/base/docs/reference/architecture.md
+1-5Lines changed: 1 addition & 5 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -45,9 +45,5 @@ Read when system shape, boundaries, or invariants matter. Keep this focused on s
45
45
- Key dependency: an independent evaluator or explicit acceptance pass writing the record as it verifies.
46
46
47
47
- Component: `.agents/skills/software-delivery/`
48
-
- Responsibility: routes non-trivial software work between feature discovery, multi-phase control, harness control, plan reviews, and independent frontend evaluation.
48
+
- Responsibility: routes non-trivial software work between feature discovery, harness control, plan reviews, and independent frontend evaluation.
49
49
- Key dependency: the request being classified by the dominant delivery need rather than by convenience.
- Purpose: router family for analytical requests across state calibration, problem framing, strategic foresight, hidden-rule reality checks, structured advisory, and multi-lens analysis.
25
21
- Update when: a reasoning leaf is added, removed, renamed, or materially repurposed.
@@ -41,7 +37,7 @@ Read when you need to find where to work. Prefer only high-value paths.
Copy file name to clipboardExpand all lines: templates/base/docs/reference/lessons.md
-4Lines changed: 0 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -17,7 +17,3 @@ Read after mistakes, rework, or surprises. Capture only reusable lessons.
17
17
- Fix:
18
18
- When to apply:
19
19
- Evidence:
20
-
21
-
- Fix: write router and skill edits under `templates/base/`, not the user-home copy, and remove any accidental repo-root `.agents/` duplicate before verifying.
22
-
- When to apply: whenever scaffolding or copying skill packages in this repository.
23
-
- Evidence: this task initially wrote to the wrong location and left a duplicate tree until the repo-root copy was deleted.
Copy file name to clipboardExpand all lines: templates/base/docs/reference/memory.md
-4Lines changed: 0 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -12,7 +12,3 @@ Read for durable truths worth preserving across sessions. Do not store transient
12
12
- Decision: `using-agent-practices` is the canonical top-level discoverability router for the template skill suite, and its `references/category-map.md` inventory must stay in sync with every live top-level skill under `templates/base/.agents/skills/`.
13
13
- Preserve because: missing standalone leaves in the top-level router create silent discovery holes even when the skills themselves are valid and present on disk.
14
14
- Revisit only if: the template adopts a different top-level discovery mechanism or moves standalone specialist skills behind a new explicit family router.
15
-
16
-
- Decision: `software-delivery/multi-phase-control` is the canonical leaf for phase-gated roadmap persistence and drift prevention, and `software-delivery` routes to it before `harness-design` when preserving original intent across phases is the primary need.
17
-
- Preserve because: multi-phase work fails when original intent lives only in chat context instead of an explicit persistence workflow.
18
-
- Revisit only if: phase preservation becomes part of another router or a new control model replaces the current multi-phase workflow.
0 commit comments