@tsfpp/agents 1.3.3 → 1.3.4
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/CHANGELOG.md
CHANGED
|
@@ -10,6 +10,15 @@ Versioning follows [Semantic Versioning](https://semver.org/).
|
|
|
10
10
|
|
|
11
11
|
## [Unreleased]
|
|
12
12
|
|
|
13
|
+
## [1.3.4] - 2026-05-18
|
|
14
|
+
|
|
15
|
+
### Changed
|
|
16
|
+
|
|
17
|
+
- Updated `audit`, `backfill-tests`, and `refactor-engineer` so they do not hand off too easily when workable slices of implementation are readily available.
|
|
18
|
+
- Updated audit/backfill log filename convention to include both time and focus.
|
|
19
|
+
- Improved handover flow from `backfill-tests` to `audit`.
|
|
20
|
+
- Added additional summary domains to report on in the audit log.
|
|
21
|
+
|
|
13
22
|
## [1.3.3] - 2026-05-17
|
|
14
23
|
|
|
15
24
|
### Added
|
|
@@ -39,13 +39,13 @@ If any referenced file is missing, stop immediately and report the path. Do not
|
|
|
39
39
|
|
|
40
40
|
## Session start
|
|
41
41
|
|
|
42
|
-
If
|
|
42
|
+
If `target` and `focus` are present in the message (e.g. `target=src/ focus=test`) or can be inferred from handoff context (e.g. previous agent worked on specific files), proceed immediately without asking.
|
|
43
|
+
|
|
44
|
+
If and only if either is missing and cannot be inferred, ask once:
|
|
43
45
|
|
|
44
46
|
> **Target** — path, package name, or layer to audit (e.g. `src/domain`, `@tsfpp/prelude`, `api layer`)?
|
|
45
47
|
> **Focus** — `all` · `types` · `boundary` · `complexity` · `loc` · `annotations` · `security` · `react` · `data` · `prelude` · `test` · or comma-separated combination?
|
|
46
48
|
|
|
47
|
-
Do not proceed until both are confirmed.
|
|
48
|
-
|
|
49
49
|
---
|
|
50
50
|
|
|
51
51
|
## Mission
|
|
@@ -59,7 +59,7 @@ Systematically inspect the target for TSF++ violations. Slice the work into mana
|
|
|
59
59
|
Create the report file **before starting any inspection**:
|
|
60
60
|
|
|
61
61
|
```
|
|
62
|
-
docs/audits/<target-slug>-<YYYYMMDD-HHmm>.md
|
|
62
|
+
docs/audits/<target-slug>-<focus>-<YYYYMMDD-HHmm>.md
|
|
63
63
|
```
|
|
64
64
|
|
|
65
65
|
Use this template exactly:
|
|
@@ -79,13 +79,20 @@ Use this template exactly:
|
|
|
79
79
|
|
|
80
80
|
> Fill in after all slices are complete.
|
|
81
81
|
|
|
82
|
-
| Category | Violations | Deviations | Passed |
|
|
83
|
-
|
|
84
|
-
| Types | — | — | — |
|
|
85
|
-
| Purity | — | — | — |
|
|
86
|
-
| Boundary | — | — | — |
|
|
87
|
-
| Annotations | — | — | — |
|
|
88
|
-
| Complexity | — | — | — |
|
|
82
|
+
| Category | Violations | Deviations | Passed | N/A |
|
|
83
|
+
|-------------|-----------|------------|--------|-----|
|
|
84
|
+
| Types | — | — | — | — |
|
|
85
|
+
| Purity | — | — | — | — |
|
|
86
|
+
| Boundary | — | — | — | — |
|
|
87
|
+
| Annotations | — | — | — | — |
|
|
88
|
+
| Complexity | — | — | — | — |
|
|
89
|
+
| Prelude | — | — | — | — |
|
|
90
|
+
| React | — | — | — | — |
|
|
91
|
+
| Data | — | — | — | — |
|
|
92
|
+
| Security | — | — | — | — |
|
|
93
|
+
| Tests | — | — | — | — |
|
|
94
|
+
|
|
95
|
+
_N/A — focus not applicable to this target (e.g. React row when no `.tsx` files in scope)_
|
|
89
96
|
|
|
90
97
|
---
|
|
91
98
|
|
|
@@ -259,7 +266,10 @@ All focus areas above in sequence. For `.tsx` files, include `react` automatical
|
|
|
259
266
|
List all files in scope. Group into logical slices (≤ 300 LOC per slice, or one cohesive module). Populate the slice index table in the report.
|
|
260
267
|
|
|
261
268
|
**Step 2 — Create report**
|
|
262
|
-
Write `docs/audits/<slug>-<
|
|
269
|
+
Write `docs/audits/<target-slug>-<focus>-<YYYYMMDD-HHmm>.md` with the template above before touching any source file.
|
|
270
|
+
Example: `docs/audits/src-domain-prelude-20260517-1430.md` or `docs/audits/src-all-20260517-0900.md`.
|
|
271
|
+
|
|
272
|
+
> **Do not suggest handoffs or pause between slices.** Work through all slices without interruption. Update the report after each slice. Only present handoff options after the final slice is complete and the summary table is filled in.
|
|
263
273
|
|
|
264
274
|
**Step 3 — Inspect slice by slice**
|
|
265
275
|
For each slice:
|
|
@@ -20,7 +20,7 @@ tools:
|
|
|
20
20
|
handoffs:
|
|
21
21
|
- label: Audit test coverage
|
|
22
22
|
agent: tsfpp-audit
|
|
23
|
-
prompt: "Audit the test files just written for TSF++ compliance. Focus: test."
|
|
23
|
+
prompt: "Audit the test files just written for TSF++ compliance. Use the same target as this backfill session. Focus: test. Do not ask for target or focus — infer from context and proceed immediately."
|
|
24
24
|
send: false
|
|
25
25
|
- label: Fix uncovered paths in implementation
|
|
26
26
|
agent: tsfpp-guarded-coding
|
|
@@ -71,6 +71,11 @@ You succeed when:
|
|
|
71
71
|
- Every branch and switch case is exercised
|
|
72
72
|
- All tests pass green
|
|
73
73
|
|
|
74
|
+
> **Do not suggest handoffs or pause between slices.** Work through all slices
|
|
75
|
+
> without interruption. Intermediate lint/typecheck runs and report updates are
|
|
76
|
+
> expected and correct. Only present handoff options after the final slice is
|
|
77
|
+
> complete and the backfill report is finished.
|
|
78
|
+
|
|
74
79
|
---
|
|
75
80
|
|
|
76
81
|
## Execution workflow
|
|
@@ -118,7 +123,8 @@ All tests must be green. If a test fails, the contract derivation was wrong —
|
|
|
118
123
|
|
|
119
124
|
**Step 5 — Backfill report**
|
|
120
125
|
|
|
121
|
-
Append a section to `docs/audits/backfill-<slug>-<
|
|
126
|
+
Append a section to `docs/audits/backfill-<target-slug>-<YYYYMMDD-HHmm>.md`:
|
|
127
|
+
Example: `docs/audits/backfill-src-domain-20260517-1430.md`.
|
|
122
128
|
|
|
123
129
|
````markdown
|
|
124
130
|
## Backfill — `<file>`
|
|
@@ -117,6 +117,8 @@ If a function exceeds 40 lines, cyclomatic complexity 10, or nesting depth 4: de
|
|
|
117
117
|
**Step 1 — Read the report**
|
|
118
118
|
Parse the audit report. Build a todo list of all open violations grouped by slice. State the slice order and open violation count, then proceed immediately — do not ask for confirmation.
|
|
119
119
|
|
|
120
|
+
> **Do not suggest handoffs or pause between slices.** Work through all violations without interruption. Intermediate lint/typecheck runs and report updates are expected and correct. Only present handoff options after every violation is resolved and the report is marked complete.
|
|
121
|
+
|
|
120
122
|
**Step 2 — Work slice by slice**
|
|
121
123
|
For each slice with open violations:
|
|
122
124
|
1. Read the file.
|