@antoneeo/kb-agentic-skill 1.7.0 → 1.8.0
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 +20 -0
- package/README.md +5 -1
- package/gemini-extension.json +1 -1
- package/package.json +1 -1
- package/skills/kb-agentic-skill/ENFORCEMENT.md +46 -0
- package/skills/kb-agentic-skill/SKILL.md +2 -2
- package/skills/kb-agentic-skill/scripts/sdlc_check.py +26 -0
- package/skills/kb-agentic-skill/scripts/sdlc_core.py +104 -0
package/CHANGELOG.md
CHANGED
|
@@ -2,6 +2,26 @@
|
|
|
2
2
|
|
|
3
3
|
Every significant change to this skill is recorded here.
|
|
4
4
|
|
|
5
|
+
## [1.8.0] - 2026-08-28
|
|
6
|
+
|
|
7
|
+
F-041 -- the mid-session drift field report, both items.
|
|
8
|
+
|
|
9
|
+
### Added
|
|
10
|
+
- **`check` notes an unwired or dead session-orientation hook.** `init` wires the
|
|
11
|
+
hook at init time only; pre-F-036 and agent-bootstrapped projects never got it,
|
|
12
|
+
silently. `check` now prints, right after its summary line, a one-line `[note]` when no hook is detectable
|
|
13
|
+
in `.claude/settings.json` / `.claude/settings.local.json` / `.codex/hooks.json`,
|
|
14
|
+
and a distinct note when a hook is wired but its validator does not resolve
|
|
15
|
+
(any-resolves aggregation across entries; BOM-tolerant reads; fail-open on
|
|
16
|
+
malformed config). Informational only: never the exit code, never a `validate`
|
|
17
|
+
warning -- CI stays green because the wired hook legitimately lives in the
|
|
18
|
+
git-ignored local settings file.
|
|
19
|
+
- **`remind` -- an opt-in per-turn one-line reminder (UserPromptSubmit).** orient
|
|
20
|
+
fires once; mid-session drift had no mechanical guard. One constant, zero-read
|
|
21
|
+
line re-arms the kb minimum (triage levels, the topics/INDEX.md recall trigger,
|
|
22
|
+
the capture moment) at every prompt. Never wired by `init`, never a default:
|
|
23
|
+
ENFORCEMENT par.4 documents the snippet and the ~50 tokens/prompt cost.
|
|
24
|
+
|
|
5
25
|
## [1.7.0] - 2026-08-28
|
|
6
26
|
|
|
7
27
|
F-040 — the capture moment (second-brain unit 2): the day's decisions reach the
|
package/README.md
CHANGED
|
@@ -82,7 +82,11 @@ it: a repo that vendors it gets a repo-relative command in the shared `.claude/s
|
|
|
82
82
|
(portable — commit it); otherwise the path is machine-specific and goes to the git-ignored
|
|
83
83
|
`.claude/settings.local.json`, and each teammate runs `init` once. Re-running `init` never
|
|
84
84
|
duplicates the hook, and a hook whose validator no longer resolves is reported as **broken**
|
|
85
|
-
with the correction rather than counted as installed.
|
|
85
|
+
with the correction rather than counted as installed. Since F-041, `check` also
|
|
86
|
+
prints a one-line note when a project has no orientation hook wired (or a dead
|
|
87
|
+
one), and an opt-in per-turn reminder exists for long sessions: `sdlc_check.py
|
|
88
|
+
remind`, one constant line via UserPromptSubmit -- documented in `ENFORCEMENT.md`
|
|
89
|
+
par.4, never wired by default.
|
|
86
90
|
|
|
87
91
|
## The family: three lenses, one spine
|
|
88
92
|
|
package/gemini-extension.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "kb-agentic-skill",
|
|
3
|
-
"version": "1.
|
|
3
|
+
"version": "1.8.0",
|
|
4
4
|
"description": "Knowledge-Base & Document-First protocol with risk triage, Vision governance, signal distillation and optional devPNT integration.",
|
|
5
5
|
"author": "Antonio Pinto (https://github.com/Antoneeo)"
|
|
6
6
|
}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@antoneeo/kb-agentic-skill",
|
|
3
|
-
"version": "1.
|
|
3
|
+
"version": "1.8.0",
|
|
4
4
|
"description": "Knowledge-Base & Document-First protocol for Claude Code, Gemini CLI, Google Antigravity and Codex with risk triage, Vision governance, signal distillation and optional devPNT integration.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"claude-code",
|
|
@@ -86,6 +86,20 @@ that cannot be placed in a command safely, or a settings file it cannot write. `
|
|
|
86
86
|
prints which case applied, per client — a silent skip is what let "documented default
|
|
87
87
|
that nobody installs" survive in the first place.
|
|
88
88
|
|
|
89
|
+
**Older and agent-bootstrapped projects are silently unwired -- `check` now tells you (F-041).**
|
|
90
|
+
`init` wires the hook at init time only: a project initialized before F-036, or whose
|
|
91
|
+
docs root was bootstrapped by the agent (which runs `index`, never `init`), never
|
|
92
|
+
received it retroactively -- and until F-041 nothing said so. `check` now prints, right after its summary line, a
|
|
93
|
+
one-line `[note]` when no orientation hook is detectable (`.claude/settings.json`,
|
|
94
|
+
`.claude/settings.local.json`, `.codex/hooks.json`), and a DISTINCT note when a hook is
|
|
95
|
+
wired but the validator it names does not resolve (the wired-and-DEAD state below --
|
|
96
|
+
silence there would bless the worst of the three states). The notes are informational
|
|
97
|
+
only: never the exit code, never a `validate` warning -- CI (`validate --strict`) does
|
|
98
|
+
not see them, because the wired hook legitimately lives in the git-ignored
|
|
99
|
+
`settings.local.json`. Known residuals, accepted as honest: a client with no hook
|
|
100
|
+
mechanism (Gemini today) keeps the note -- read it as the manual-reads reminder; a
|
|
101
|
+
`.codex/hooks.json` wired in a shape other than the documented one keeps it too.
|
|
102
|
+
|
|
89
103
|
**Which settings file, and why it is not always the shared one.** The command names a
|
|
90
104
|
validator, and where that validator lives decides where the hook may be written:
|
|
91
105
|
|
|
@@ -133,6 +147,38 @@ Gemini CLI — wire the same command into its startup-hook mechanism if present;
|
|
|
133
147
|
- **Hybrid/devPNT projects**: add `--hybrid` — the hook then appends a one-line pointer to run `devpnt_mcp_get_bootstrap` for the Master Plan / Knowledge Layer and does not replicate them; the filesystem orientation (router + handoff + README) still emits.
|
|
134
148
|
- Like the CI gate (§2), if you copied `sdlc_check.py` into the repo, the hook references that copy — keep it current when you update the skill.
|
|
135
149
|
|
|
150
|
+
### Per-turn reminder: `remind` (UserPromptSubmit -- opt-in, NEVER a default)
|
|
151
|
+
|
|
152
|
+
`orient` fires once, at session start; a long session drifts after it (a compaction,
|
|
153
|
+
ten turns of other work -- the F-041 field report). `sdlc_check.py remind` prints ONE
|
|
154
|
+
constant line -- the kb minimum: the triage levels, the topic-index recall trigger, the
|
|
155
|
+
capture moment -- for the client to inject at every prompt. Cost: ~50 tokens per
|
|
156
|
+
prompt, which is exactly why `init` never wires it and never will by default: the
|
|
157
|
+
per-turn cost is the project's choice. The line is constant and zero-read, so nothing
|
|
158
|
+
repo-controlled can ride into the agent's context through it, and the vision's
|
|
159
|
+
no-per-turn-graph-queries non-goal stays honored -- the line re-arms the
|
|
160
|
+
entry-into-topic trigger; it never runs the reading itself.
|
|
161
|
+
|
|
162
|
+
Claude Code -- in the project's `.claude/settings.json` (or `settings.local.json` when
|
|
163
|
+
the validator path is machine-specific, same rule as the orientation hook above):
|
|
164
|
+
|
|
165
|
+
```json
|
|
166
|
+
{
|
|
167
|
+
"hooks": {
|
|
168
|
+
"UserPromptSubmit": [
|
|
169
|
+
{
|
|
170
|
+
"hooks": [
|
|
171
|
+
{
|
|
172
|
+
"type": "command",
|
|
173
|
+
"command": "python \"C:\\Users\\<user>\\.claude\\skills\\kb-agentic\\scripts\\sdlc_check.py\" remind"
|
|
174
|
+
}
|
|
175
|
+
]
|
|
176
|
+
}
|
|
177
|
+
]
|
|
178
|
+
}
|
|
179
|
+
}
|
|
180
|
+
```
|
|
181
|
+
|
|
136
182
|
## 5. Skill eval battery (release gate)
|
|
137
183
|
|
|
138
184
|
**Skill development only.** `test_*.py` and `evals/` are deliberately absent from the npm `files` allowlist — they never reach an installed consumer, so this section applies to whoever builds the skill, not to a project that uses it. (Consumers get `sdlc_check.py`; §1–§4 are theirs.)
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: kb-agentic
|
|
3
|
-
version: 1.
|
|
3
|
+
version: 1.8.0
|
|
4
4
|
description: Knowledge-Base & Document-First protocol with risk-proportional triage, Vision as a guide, Signal Distillation, a complete Standalone mode and optional symbiosis with devPNT. Use for user documentation, knowledge extraction, SOPs, research notes, decision logs, knowledge management — and for answering project questions from the claim ledger and topic graph.
|
|
5
5
|
author: Antonio Pinto (https://github.com/Antoneeo)
|
|
6
6
|
copyright: (c) 2026 Antonio Pinto
|
|
@@ -23,7 +23,7 @@ Support files in the skill directory:
|
|
|
23
23
|
- `review.md`: the review discipline — when a review is due, how to request one, how to receive findings, how to review.
|
|
24
24
|
- `dispatch.md`: opt-in subagent execution of an approved plan.
|
|
25
25
|
- `routing.md`: which lens owns this unit of work. Read ONLY when a sibling lens skill is installed alongside this one; a single-lens install never reads it.
|
|
26
|
-
- `scripts/sdlc_check.py` + `scripts/sdlc_core.py`: the mechanical validator for the docs root (`check`, `validate`, `index`, `stale`, `mark`, `gate`, `plan`, `orient`, `migrate`, and the knowledge overlay: `graph`, `corpus`, `claim-id`, `anchor`, `export`, `import`). Two files: the core is the family's shared spine; the entry point IS the knowledge overlay — the claim ledger and topic-graph checks live inside it, so the core alone runs none of them. Copy both, or neither.
|
|
26
|
+
- `scripts/sdlc_check.py` + `scripts/sdlc_core.py`: the mechanical validator for the docs root (`check`, `validate`, `index`, `stale`, `mark`, `gate`, `plan`, `orient`, `migrate`, and the knowledge overlay: `graph`, `corpus`, `claim-id`, `anchor`, `export`, `import`, plus `remind` — the opt-in per-turn one-line reminder, ENFORCEMENT §4). Two files: the core is the family's shared spine; the entry point IS the knowledge overlay — the claim ledger and topic-graph checks live inside it, so the core alone runs none of them. Copy both, or neither.
|
|
27
27
|
- `ENFORCEMENT.md`: optional setup for CI and hooks.
|
|
28
28
|
|
|
29
29
|
Read these files only when needed. `SKILL.md` is the operating contract; the support files are progressive resources.
|
|
@@ -1598,6 +1598,26 @@ def kb_cmd_orient(argv):
|
|
|
1598
1598
|
return rc
|
|
1599
1599
|
|
|
1600
1600
|
|
|
1601
|
+
def kb_cmd_remind(_argv):
|
|
1602
|
+
"""F-041 item B: the per-turn one-line reminder (UserPromptSubmit, opt-in).
|
|
1603
|
+
|
|
1604
|
+
A CONSTANT, zero-read line -- the kb minimum an agent needs re-armed every
|
|
1605
|
+
prompt: triage, the recall trigger, the capture moment. Constant by
|
|
1606
|
+
contract: nothing repo-controlled may ride a line injected into the agent's
|
|
1607
|
+
context every turn, and zero reads is what keeps the per-turn cost flat.
|
|
1608
|
+
The vision's non-goal bans per-turn GRAPH queries; this line only re-arms
|
|
1609
|
+
the entry-into-topic criterion, it never runs the descent. Never wired by
|
|
1610
|
+
init and never a default: ENFORCEMENT par.4 documents the manual snippet
|
|
1611
|
+
and the ~50 tokens/prompt cost -- the project chooses. Fail-open like every
|
|
1612
|
+
hook surface: argv is ignored whole, the exit code is always 0."""
|
|
1613
|
+
print("kb-agentic: triage before acting -- L1 quick fact, L2 propagation of "
|
|
1614
|
+
"settled knowledge, L3 new knowledge unit, Spike exploration; declare "
|
|
1615
|
+
"the level. Answering about this project's domain? Read topics/INDEX.md "
|
|
1616
|
+
"under the docs root first and cite claim ids, or state no coverage. "
|
|
1617
|
+
"Decisions made today get recorded in the KB at session close.")
|
|
1618
|
+
return 0
|
|
1619
|
+
|
|
1620
|
+
|
|
1601
1621
|
def kb_cmd_help():
|
|
1602
1622
|
"""The spine's usage, then the overlay's own commands.
|
|
1603
1623
|
|
|
@@ -1619,6 +1639,8 @@ knowledge overlay (kb-agentic) -- also available:
|
|
|
1619
1639
|
anchor <path> <phrase> resolve a phrase to a verified locator span
|
|
1620
1640
|
export --out <dir> bundle a subgraph WITH the bytes its claims cite
|
|
1621
1641
|
import <dir> merge a bundle in additively (never overwrites)
|
|
1642
|
+
remind one-line per-turn reminder (UserPromptSubmit hook,
|
|
1643
|
+
opt-in -- ENFORCEMENT par.4; constant, zero-read)
|
|
1622
1644
|
|
|
1623
1645
|
index / validate / check the spine's behaviour PLUS the claim ledger and
|
|
1624
1646
|
the topic graph""")
|
|
@@ -1636,6 +1658,10 @@ def main(argv=None):
|
|
|
1636
1658
|
# spine runs it on the raw argv, the overlay only appends the topic router.
|
|
1637
1659
|
if argv and argv[0] == "orient":
|
|
1638
1660
|
return kb_cmd_orient(argv)
|
|
1661
|
+
# `remind` too: it is a hook surface (UserPromptSubmit), so a stray flag
|
|
1662
|
+
# must print the line, not an argparse usage error into someone's prompt.
|
|
1663
|
+
if argv and argv[0] == "remind":
|
|
1664
|
+
return kb_cmd_remind(argv)
|
|
1639
1665
|
# Forward-by-default: anything not intercepted goes to the spine untouched.
|
|
1640
1666
|
# Never a hand-copied command tuple - that is how a spine command gets
|
|
1641
1667
|
# silently dropped (mkt_check.py ships that exact defect with `migrate`).
|
|
@@ -1646,6 +1646,109 @@ def cmd_mark(root, paths):
|
|
|
1646
1646
|
return 0
|
|
1647
1647
|
|
|
1648
1648
|
|
|
1649
|
+
# ------------------------------------------------- orientation-hook wiring
|
|
1650
|
+
# F-041: read-only mirror of the installer-side detection in lib.js
|
|
1651
|
+
# (findOrientHook + orientHookValidator). The JS owns WRITING hooks, with its
|
|
1652
|
+
# injection guards; this side only DETECTS, at `check`, so a project whose
|
|
1653
|
+
# sessions start unoriented hears about it in-project (ENFORCEMENT par.4: the
|
|
1654
|
+
# hook is wired at init only -- pre-F-036 and agent-bootstrapped projects never
|
|
1655
|
+
# got it, and a wired-but-dead hook is "the worst of the three states").
|
|
1656
|
+
# The file list is a SUPERSET of the JS one: .codex/hooks.json is documented in
|
|
1657
|
+
# ENFORCEMENT par.4 but never written by init. Entry-point names are pinned
|
|
1658
|
+
# equal on both sides.
|
|
1659
|
+
ORIENT_HOOK_ENTRY_POINTS = ("sdlc_check.py", "mkt_check.py")
|
|
1660
|
+
ORIENT_HOOK_CONFIG_FILES = (".claude/settings.json",
|
|
1661
|
+
".claude/settings.local.json",
|
|
1662
|
+
".codex/hooks.json")
|
|
1663
|
+
ORIENT_HOOK_CONFIG_CAP = 1024 * 1024 # settings are tiny; cap pathological files
|
|
1664
|
+
_ORIENT_HOOK_CMD_RE = re.compile(r" orient(\s|$)")
|
|
1665
|
+
_ORIENT_HOOK_TOKEN_RE = re.compile(r'"([^"]*)"|(\S+)')
|
|
1666
|
+
|
|
1667
|
+
|
|
1668
|
+
def _orient_hook_commands(root):
|
|
1669
|
+
"""Yield the command strings of SessionStart entries invoking an orient
|
|
1670
|
+
entry point, across every known client config under root. Fail-open: an
|
|
1671
|
+
unreadable, oversized, malformed or directory-shaped file contributes
|
|
1672
|
+
nothing. Reads are utf-8-sig -- a BOM'd settings.json (PowerShell/Notepad)
|
|
1673
|
+
must not read as unwired (declared divergence from Node's JSON.parse)."""
|
|
1674
|
+
for rel in ORIENT_HOOK_CONFIG_FILES:
|
|
1675
|
+
p = Path(root) / rel
|
|
1676
|
+
try:
|
|
1677
|
+
if not p.is_file() or p.stat().st_size > ORIENT_HOOK_CONFIG_CAP:
|
|
1678
|
+
continue
|
|
1679
|
+
data = json.loads(p.read_text(encoding="utf-8-sig"))
|
|
1680
|
+
except Exception:
|
|
1681
|
+
continue
|
|
1682
|
+
if not isinstance(data, dict):
|
|
1683
|
+
continue
|
|
1684
|
+
hooks = data.get("hooks")
|
|
1685
|
+
groups = hooks.get("SessionStart") if isinstance(hooks, dict) else None
|
|
1686
|
+
if groups is None:
|
|
1687
|
+
groups = data.get("SessionStart") # the .codex/hooks.json shape
|
|
1688
|
+
if not isinstance(groups, list):
|
|
1689
|
+
continue
|
|
1690
|
+
for group in groups:
|
|
1691
|
+
if not isinstance(group, dict):
|
|
1692
|
+
continue
|
|
1693
|
+
inner = group.get("hooks")
|
|
1694
|
+
for h in inner if isinstance(inner, list) else [group]:
|
|
1695
|
+
cmd = h.get("command") if isinstance(h, dict) else None
|
|
1696
|
+
if (isinstance(cmd, str)
|
|
1697
|
+
and any(n in cmd for n in ORIENT_HOOK_ENTRY_POINTS)
|
|
1698
|
+
and _ORIENT_HOOK_CMD_RE.search(cmd)):
|
|
1699
|
+
yield cmd
|
|
1700
|
+
|
|
1701
|
+
|
|
1702
|
+
def orient_hook_state(root):
|
|
1703
|
+
"""'wired' | 'dead' | 'unwired'.
|
|
1704
|
+
|
|
1705
|
+
A matching entry whose tokens name no validator is "cannot tell" and counts
|
|
1706
|
+
as wired (the lib.js contract: never report broken on cannot-tell). 'dead'
|
|
1707
|
+
requires every matching entry to name a validator that does not resolve:
|
|
1708
|
+
the client runs ALL SessionStart hooks, so one live entry beside a dead one
|
|
1709
|
+
still orients the session (any-resolves aggregation). Relative tokens
|
|
1710
|
+
resolve against the project root -- the vendored-validator convention."""
|
|
1711
|
+
matched = False
|
|
1712
|
+
for cmd in _orient_hook_commands(root):
|
|
1713
|
+
matched = True
|
|
1714
|
+
token = None
|
|
1715
|
+
for m in _ORIENT_HOOK_TOKEN_RE.finditer(cmd):
|
|
1716
|
+
tok = m.group(1) if m.group(1) is not None else m.group(2)
|
|
1717
|
+
if any(tok.endswith(n) for n in ORIENT_HOOK_ENTRY_POINTS):
|
|
1718
|
+
token = tok
|
|
1719
|
+
break
|
|
1720
|
+
if token is None:
|
|
1721
|
+
return "wired"
|
|
1722
|
+
tp = Path(token)
|
|
1723
|
+
if not tp.is_absolute():
|
|
1724
|
+
tp = Path(root) / tp
|
|
1725
|
+
if tp.exists():
|
|
1726
|
+
return "wired"
|
|
1727
|
+
return "dead" if matched else "unwired"
|
|
1728
|
+
|
|
1729
|
+
|
|
1730
|
+
def print_orient_hook_note(root):
|
|
1731
|
+
"""Print the FS-A wiring note (unwired / dead), or nothing when wired.
|
|
1732
|
+
|
|
1733
|
+
Informational only, by contract: never an exit code, never a validate
|
|
1734
|
+
warning (--strict CI must not fail on a hook that legitimately lives in
|
|
1735
|
+
git-ignored settings.local.json). Called by the spine's cmd_check AND by
|
|
1736
|
+
an overlay that replaces check with its own pipeline (the marketing lens
|
|
1737
|
+
does): the note is check-layer behaviour, whoever owns the check."""
|
|
1738
|
+
try:
|
|
1739
|
+
state = orient_hook_state(root)
|
|
1740
|
+
if state == "unwired":
|
|
1741
|
+
print("[note] no session-orientation hook wired (.claude/settings.json, "
|
|
1742
|
+
".claude/settings.local.json, .codex/hooks.json): sessions start "
|
|
1743
|
+
"unoriented -- run the installer's init or wire it per ENFORCEMENT.md par.4")
|
|
1744
|
+
elif state == "dead":
|
|
1745
|
+
print("[note] a session-orientation hook is wired but the validator it names "
|
|
1746
|
+
"does not resolve: sessions start unoriented -- re-run init or fix the "
|
|
1747
|
+
"command (ENFORCEMENT.md par.4)")
|
|
1748
|
+
except Exception:
|
|
1749
|
+
pass # detection must never break check
|
|
1750
|
+
|
|
1751
|
+
|
|
1649
1752
|
def cmd_check(root, strict=False, hybrid=False):
|
|
1650
1753
|
print("===== validate =====")
|
|
1651
1754
|
rc_v = cmd_validate(root, strict=strict, hybrid=hybrid)
|
|
@@ -1653,6 +1756,7 @@ def cmd_check(root, strict=False, hybrid=False):
|
|
|
1653
1756
|
rc_s = cmd_stale(root, hybrid=hybrid)
|
|
1654
1757
|
print(f"\ncheck: {'CLEAN' if not (rc_v or rc_s) else 'NOT CLEAN'} "
|
|
1655
1758
|
f"(validate rc={rc_v}, stale rc={rc_s})")
|
|
1759
|
+
print_orient_hook_note(root)
|
|
1656
1760
|
return 1 if (rc_v or rc_s) else 0
|
|
1657
1761
|
|
|
1658
1762
|
|