djournal 0.12.1 → 0.13.0
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/.agents/adapters/contract.md +5 -0
- package/.agents/adapters/shared/journal-hook.js +2 -2
- package/.agents/rules/AUTOMATION.md +30 -7
- package/.agents/rules/FEAT.md +1 -1
- package/.agents/rules/STATE.md +3 -3
- package/.agents/skills/{audit → journal-audit}/SKILL.md +1 -1
- package/.agents/skills/journal-audit/agents/openai.yaml +4 -0
- package/.agents/skills/{decision → journal-decision}/SKILL.md +1 -1
- package/.agents/skills/journal-decision/agents/openai.yaml +4 -0
- package/.agents/skills/{document → journal-document}/SKILL.md +2 -2
- package/.agents/skills/{journal → journal-entry}/SKILL.md +2 -2
- package/.agents/skills/{init-work → journal-init-work}/SKILL.md +1 -1
- package/.agents/skills/{plan → journal-plan}/SKILL.md +9 -3
- package/.agents/skills/journal-plan/agents/openai.yaml +4 -0
- package/.agents/skills/{recall → journal-recall}/SKILL.md +9 -4
- package/.agents/skills/journal-recall/agents/openai.yaml +4 -0
- package/.agents/skills/{reconcile → journal-reconcile}/SKILL.md +3 -3
- package/.agents/skills/journal-reconcile/agents/openai.yaml +4 -0
- package/.agents/skills/{reinforce → journal-reinforce}/SKILL.md +4 -4
- package/.agents/skills/{research-codebase → journal-research-codebase}/SKILL.md +1 -1
- package/.agents/skills/{research-web → journal-research-web}/SKILL.md +1 -1
- package/.agents/skills/{resume → journal-resume}/SKILL.md +8 -3
- package/.agents/skills/journal-resume/agents/openai.yaml +4 -0
- package/.agents/skills/{switch → journal-switch}/SKILL.md +3 -3
- package/.agents/skills/journal-workflow/SKILL.md +31 -12
- package/AGENTS.md +11 -2
- package/README.md +5 -1
- package/docs/installation.md +48 -0
- package/docs/read-only-team-journal.md +2 -2
- package/docs/recall.md +7 -2
- package/docs/remote-sync.md +1 -1
- package/integrations/grok-bot/skill-registration.md +17 -17
- package/lib/installer/index.js +1 -1
- package/lib/setup.js +1 -1
- package/package.json +1 -1
- package/spec.md +26 -24
- package/.agents/skills/audit/agents/openai.yaml +0 -4
- package/.agents/skills/decision/agents/openai.yaml +0 -4
- package/.agents/skills/reconcile/agents/openai.yaml +0 -4
|
@@ -16,6 +16,11 @@ second workflow engine and do not own durable state.
|
|
|
16
16
|
An adapter may omit events its harness cannot support reliably. Instruction-only
|
|
17
17
|
behavior remains the compatibility baseline.
|
|
18
18
|
|
|
19
|
+
Start/prompt reminders name the `journal-*` skills and point to the invocation
|
|
20
|
+
contract in `AUTOMATION.md`. They do not infer which skill a request needs,
|
|
21
|
+
track skill calls, or block native memory or planning tools. Closure validation
|
|
22
|
+
does not prove that resume, recall, or planning happened earlier.
|
|
23
|
+
|
|
19
24
|
Grok Bot is an instruction and skill-only harness. It has no SessionStart or
|
|
20
25
|
Stop hooks, so it omits every semantic event. Compliance is the installed
|
|
21
26
|
skills, `AGENTS.md`, and the `djournal` / `journal` CLI. The recipe lives at
|
|
@@ -303,7 +303,7 @@ function handle(payload, options = {}) {
|
|
|
303
303
|
? updateNotice(runUpdateCheck(root, options.updateRunner))
|
|
304
304
|
: "";
|
|
305
305
|
const suffix = selectionSuffix(root, payload, env);
|
|
306
|
-
return context(event, `Follow AGENTS.md and .agents/rules/AUTOMATION.md.${pullMessage}${updateMessage}${suffix}
|
|
306
|
+
return context(event, `Follow AGENTS.md and .agents/rules/AUTOMATION.md.${pullMessage}${updateMessage}${suffix} When project context is needed, load journal-resume; for prior decisions/history, load journal-recall; before work meeting the planning threshold, load journal-plan. Invoke the exact skill name, or read .agents/skills/<skill-name>/SKILL.md if unavailable. Native memory, conversation resume, and plan mode do not replace these journal steps. Honor journal: off.`);
|
|
307
307
|
}
|
|
308
308
|
|
|
309
309
|
if (event === "UserPromptSubmit") {
|
|
@@ -318,7 +318,7 @@ function handle(payload, options = {}) {
|
|
|
318
318
|
: `The requested journal pull failed and must be resolved before publication: ${result.message}`;
|
|
319
319
|
return context(event, `${pullMessage} Classify this request using the ambient journal workflow and run journal sync if publication was requested.`);
|
|
320
320
|
}
|
|
321
|
-
return context(event, "Classify this request using the ambient journal workflow. Close meaningful changed state before the final response, without adding ceremony to read-only or trivial work.");
|
|
321
|
+
return context(event, "Classify this request using the ambient journal workflow. Load journal-workflow and the selected journal-* skills when required; invoke the skill or read its .agents/skills/<skill-name>/SKILL.md. Close meaningful changed state before the final response, without adding ceremony to read-only or trivial work.");
|
|
322
322
|
}
|
|
323
323
|
|
|
324
324
|
if (event !== "Stop") return {};
|
|
@@ -17,20 +17,43 @@ Classify each request before acting:
|
|
|
17
17
|
Classification requires semantic judgment. File changes alone do not establish
|
|
18
18
|
that a task was meaningful.
|
|
19
19
|
|
|
20
|
+
## Skill invocation
|
|
21
|
+
|
|
22
|
+
When a checkpoint calls for a skill, load it by its exact `journal-*` name
|
|
23
|
+
through the harness skill tool. If that mechanism is unavailable or the skill
|
|
24
|
+
is not listed, read `.agents/skills/<skill-name>/SKILL.md` and follow it. Naming
|
|
25
|
+
a skill in a response, recalling a conversation, or entering native plan mode
|
|
26
|
+
does not execute its procedure. If neither the skill nor its file is available,
|
|
27
|
+
report the missing installation instead of claiming it ran.
|
|
28
|
+
|
|
29
|
+
Use these routes when the request needs project memory:
|
|
30
|
+
|
|
31
|
+
| Intent | Skill | Required evidence or output |
|
|
32
|
+
| --- | --- | --- |
|
|
33
|
+
| Continue work, recover state after compaction, find next steps | `journal-resume` | Session-selected work and latest canonical spine entry |
|
|
34
|
+
| Explain an earlier choice, retrieve project history or related evidence | `journal-recall` | Relevant canonical entries and source citations |
|
|
35
|
+
| Design work meeting the plan threshold below | `journal-plan` | Canonical plan before implementation, or an existing applicable plan |
|
|
36
|
+
|
|
37
|
+
Native memory and session summaries can help locate evidence; they do not
|
|
38
|
+
replace canonical journal reads. Do not rerun a skill or reload unchanged
|
|
39
|
+
context already available for this task. Keep the opt-out and trivial-work
|
|
40
|
+
exceptions below. A harness read-only/plan mode still forbids writes: defer
|
|
41
|
+
persisting a journal plan until writes are allowed, and state that limitation.
|
|
42
|
+
|
|
20
43
|
## Start checkpoint
|
|
21
44
|
|
|
22
45
|
- If the user says `journal: off`, skip ambient journal reads and writes for the
|
|
23
46
|
request unless another explicit instruction requires them.
|
|
24
|
-
- If the resolved journal root's `state.json` selects active work, use `resume`
|
|
47
|
+
- If the resolved journal root's `state.json` selects active work, use `journal-resume`
|
|
25
48
|
when historical context can affect the result.
|
|
26
49
|
- If no active work exists and the request clearly starts a new durable work
|
|
27
|
-
stream, use `init-work`. Ask only when its identity or scope is materially
|
|
50
|
+
stream, use `journal-init-work`. Ask only when its identity or scope is materially
|
|
28
51
|
ambiguous.
|
|
29
52
|
- Do not create a work item for a read-only question or incidental edit.
|
|
30
53
|
|
|
31
54
|
## Plan threshold
|
|
32
55
|
|
|
33
|
-
Use `plan` before implementation when the work has multiple dependent phases,
|
|
56
|
+
Use `journal-plan` before implementation when the work has multiple dependent phases,
|
|
34
57
|
meaningful design choices, risky migration behavior, cross-component effects,
|
|
35
58
|
or a validation strategy that should survive the current conversation.
|
|
36
59
|
|
|
@@ -39,10 +62,10 @@ validate. Existing user-approved plans do not need another plan.
|
|
|
39
62
|
|
|
40
63
|
## Work checkpoint
|
|
41
64
|
|
|
42
|
-
- Use `research-codebase` or `research-web` only when the findings should remain
|
|
65
|
+
- Use `journal-research-codebase` or `journal-research-web` only when the findings should remain
|
|
43
66
|
durable evidence.
|
|
44
|
-
- Use `decision` when an accepted choice and its rationale constrain later work.
|
|
45
|
-
- Use `document` for a durable synthesized reference.
|
|
67
|
+
- Use `journal-decision` when an accepted choice and its rationale constrain later work.
|
|
68
|
+
- Use `journal-document` for a durable synthesized reference.
|
|
46
69
|
- Never copy semantic write logic into an orchestration hook or adapter.
|
|
47
70
|
|
|
48
71
|
## Closure checkpoint
|
|
@@ -50,7 +73,7 @@ validate. Existing user-approved plans do not need another plan.
|
|
|
50
73
|
Before the final response:
|
|
51
74
|
|
|
52
75
|
1. Decide whether meaningful journal state changed.
|
|
53
|
-
2. If it did, use `journal` once after validation and reference supporting
|
|
76
|
+
2. If it did, use `journal-entry` once after validation and reference supporting
|
|
54
77
|
entries actually used.
|
|
55
78
|
3. Check the latest spine entry before writing. Update or skip when a new entry
|
|
56
79
|
would describe the same completed outcome.
|
package/.agents/rules/FEAT.md
CHANGED
|
@@ -43,7 +43,7 @@ Example: `2026-06-29-01-portable-journal-skill-improvements`.
|
|
|
43
43
|
|
|
44
44
|
## Initialization
|
|
45
45
|
|
|
46
|
-
`init-work` MUST:
|
|
46
|
+
`journal-init-work` MUST:
|
|
47
47
|
|
|
48
48
|
1. Inspect existing work folders to select a non-conflicting daily sequence.
|
|
49
49
|
2. Create the complete directory structure above.
|
package/.agents/rules/STATE.md
CHANGED
|
@@ -43,9 +43,9 @@ Session binding does not change this file and is not durable journal history.
|
|
|
43
43
|
|
|
44
44
|
## Missing or invalid state
|
|
45
45
|
|
|
46
|
-
- `init-work` may create `<journal-root>/state.json` after creating the work
|
|
46
|
+
- `journal-init-work` may create `<journal-root>/state.json` after creating the work
|
|
47
47
|
folder and `work.md`.
|
|
48
|
-
- `switch` may repair state only after the user selects an existing work item.
|
|
49
|
-
- Other write skills MUST stop with: `No active work found. Run init-work or switch first.`
|
|
48
|
+
- `journal-switch` may repair state only after the user selects an existing work item.
|
|
49
|
+
- Other write skills MUST stop with: `No active work found. Run journal-init-work or journal-switch first.`
|
|
50
50
|
- Read-only recall across all work items may proceed without active state.
|
|
51
51
|
- Never guess the active work item from recency when a write would result.
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: audit
|
|
2
|
+
name: journal-audit
|
|
3
3
|
description: Inspect filesystem journal integrity without making changes. Use to report malformed or legacy metadata, broken links, misclassified entries, buried decisions, orphan supporting files, weak summaries, wrong work status, duplicate numbering, timeline drift, or safety concerns.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: decision
|
|
2
|
+
name: journal-decision
|
|
3
3
|
description: Record a clear technical or product decision as a standalone canonical supporting entry and link its evidence. Use when an option has been chosen, rationale must be preserved, or a prior decision is explicitly superseded.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: document
|
|
2
|
+
name: journal-document
|
|
3
3
|
description: Synthesize existing journal, codebase research, and web research into a concise canonical supporting document. Use when the user asks for a durable technical decision aid, how-to, walkthrough, tutorial, or architecture reference.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -28,7 +28,7 @@ Distill and connect evidence; do not concatenate source material.
|
|
|
28
28
|
- Reconcile codebase reality with external findings.
|
|
29
29
|
- Retain source attribution and note unresolved contradictions.
|
|
30
30
|
- Produce a standalone reference shorter than its inputs.
|
|
31
|
-
- Do not turn a decision aid into a recorded decision; use `decision` after an
|
|
31
|
+
- Do not turn a decision aid into a recorded decision; use `journal-decision` after an
|
|
32
32
|
option is actually chosen.
|
|
33
33
|
|
|
34
34
|
## Create the supporting entry
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: journal
|
|
2
|
+
name: journal-entry
|
|
3
3
|
description: Write a canonical journal entry for meaningful work completed in the current session. Use at the end of implementation, investigation, or a material status update.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -102,6 +102,6 @@ diffs, transcripts, environment dumps, secrets, or unbounded command output.
|
|
|
102
102
|
|
|
103
103
|
- Document only the current logical session.
|
|
104
104
|
- Do not claim tests passed unless they were run successfully.
|
|
105
|
-
- Do not create standalone decision entries implicitly; use `decision` when the
|
|
105
|
+
- Do not create standalone decision entries implicitly; use `journal-decision` when the
|
|
106
106
|
rationale deserves durable supporting context.
|
|
107
107
|
- Keep the entry concise and evidence-backed.
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: plan
|
|
3
|
-
description: Design
|
|
2
|
+
name: journal-plan
|
|
3
|
+
description: Design and persist a djournal plan before multi-phase features, risky fixes, migrations, or changes with cross-component decisions. Use when implementation needs a durable plan; a chat checklist or native plan mode alone is not a journal plan. Skip trivial changes and reuse an applicable accepted plan.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
6
|
# Plan work
|
|
@@ -8,6 +8,12 @@ description: Design a concrete unit of work and write a canonical plan entry bef
|
|
|
8
8
|
This skill produces a plan entry only. Do not implement source changes while
|
|
9
9
|
using it.
|
|
10
10
|
|
|
11
|
+
A native planning tool may help develop the design. The durable output of this
|
|
12
|
+
skill is the canonical journal plan linked to its work item. If the harness is
|
|
13
|
+
in a read-only mode, prepare the proposal and defer the write until permitted;
|
|
14
|
+
do not claim the journal plan was saved. Reuse an applicable accepted journal
|
|
15
|
+
plan instead of creating a duplicate.
|
|
16
|
+
|
|
11
17
|
## Load context
|
|
12
18
|
|
|
13
19
|
1. Read completely:
|
|
@@ -21,7 +27,7 @@ using it.
|
|
|
21
27
|
active work is unavailable.
|
|
22
28
|
3. Read the latest spine entry when present. Its timeline is the primary index.
|
|
23
29
|
4. List `_research/`, `docs/`, and `decisions/`; read only files relevant to the requested plan.
|
|
24
|
-
5. If essential codebase context is absent, invoke `research-codebase` first.
|
|
30
|
+
5. If essential codebase context is absent, invoke `journal-research-codebase` first.
|
|
25
31
|
Use sub-agents only when the current collaboration policy and user allow it.
|
|
26
32
|
|
|
27
33
|
## Analyze
|
|
@@ -1,12 +1,17 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: recall
|
|
3
|
-
description:
|
|
2
|
+
name: journal-recall
|
|
3
|
+
description: Search djournal project history and cite canonical Markdown when asked why a decision was made, what was tried, what changed, or what prior work constrains a task. Use for project history, rationale, onboarding, and evidence lookup even when native agent memory has a possible answer. Read-only.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
6
|
# Recall journal history
|
|
7
7
|
|
|
8
8
|
Answer from journal evidence. Do not modify files or invent missing history.
|
|
9
9
|
|
|
10
|
+
Use canonical journal entries as sources. Native memory and conversation history
|
|
11
|
+
may suggest search terms, but do not substitute for retrieving the journal
|
|
12
|
+
evidence. A CLI search locates sources; read the relevant entries before drawing
|
|
13
|
+
conclusions or claiming the recall procedure is complete.
|
|
14
|
+
|
|
10
15
|
## Load contracts
|
|
11
16
|
|
|
12
17
|
Read `.agents/rules/METADATA.md`, `LINKS.md`, `SAFETY.md`, `JOURNAL.md`, and
|
|
@@ -103,5 +108,5 @@ work, add a brief `Reinforcement` note before the Sources table:
|
|
|
103
108
|
steps, or the user asked to connect it to current work.
|
|
104
109
|
- `ambiguous`: list candidate source paths and say what target is unclear.
|
|
105
110
|
|
|
106
|
-
The surrounding workflow, not recall, decides whether to call `journal`,
|
|
107
|
-
`decision`, or `reinforce` before the final response.
|
|
111
|
+
The surrounding workflow, not recall, decides whether to call `journal-entry`,
|
|
112
|
+
`journal-decision`, or `journal-reinforce` before the final response.
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: reconcile
|
|
2
|
+
name: journal-reconcile
|
|
3
3
|
description: Audit and conservatively repair clear filesystem journal discrepancies. Use to add legacy frontmatter incrementally, repair deterministic metadata or links, extract explicit decisions, improve weak summaries, or correct unambiguous work status while preserving history.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -29,7 +29,7 @@ Flags may be combined. Never infer an ambiguous scope for writes.
|
|
|
29
29
|
- `.agents/rules/LINKS.md`
|
|
30
30
|
- `.agents/rules/SAFETY.md`
|
|
31
31
|
- `.agents/rules/STATE.md`
|
|
32
|
-
- `.agents/skills/audit/SKILL.md`
|
|
32
|
+
- `.agents/skills/journal-audit/SKILL.md`
|
|
33
33
|
2. Perform the audit procedure over the complete scope before changing files.
|
|
34
34
|
3. Present each proposed fix with path, fields, evidence, and reason.
|
|
35
35
|
|
|
@@ -51,7 +51,7 @@ Apply a fix only when its result is unique and supported by the contracts:
|
|
|
51
51
|
- Add a missing link when the body explicitly cites one uniquely resolved target
|
|
52
52
|
and relation semantics are clear.
|
|
53
53
|
- Link an orphan supporting entry when one unique owning entry explicitly uses it.
|
|
54
|
-
- Extract an explicit buried decision through the `decision` workflow, preserve
|
|
54
|
+
- Extract an explicit buried decision through the `journal-decision` workflow, preserve
|
|
55
55
|
the original body, and add a concise pointer only when useful.
|
|
56
56
|
- Set work status to `completed` only when the recorded outcome is unequivocally
|
|
57
57
|
complete; otherwise flag it.
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: reinforce
|
|
2
|
+
name: journal-reinforce
|
|
3
3
|
description: "Link recalled journal evidence into current or already-closed work. Use when recall found material sources that should reinforce, clarify, contradict, or supersede an existing journal entry, especially after recall ended read-only with `journal-status: not-needed` or when the user explicitly asks to update entries with recalled evidence."
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -49,12 +49,12 @@ follow-up entry would be misleading.
|
|
|
49
49
|
|
|
50
50
|
## Choose the mutation
|
|
51
51
|
|
|
52
|
-
- Current in-progress closure: let `journal` create the entry and include the
|
|
52
|
+
- Current in-progress closure: let `journal-entry` create the entry and include the
|
|
53
53
|
recalled source links.
|
|
54
54
|
- Explicit link-only update: add frontmatter links to the target when source and
|
|
55
55
|
target resolve uniquely and the relation is clear.
|
|
56
56
|
- Already-closed entry with substantive correction: create a follow-up `status`
|
|
57
|
-
entry, or use `decision` when a decision is accepted or superseded.
|
|
57
|
+
entry, or use `journal-decision` when a decision is accepted or superseded.
|
|
58
58
|
- Closed body edits: avoid them. Do only narrow, explicit edits requested by the
|
|
59
59
|
user, and never rewrite wholesale.
|
|
60
60
|
|
|
@@ -79,7 +79,7 @@ For link-only edits:
|
|
|
79
79
|
3. Update `updatedAt`.
|
|
80
80
|
4. Leave the body unchanged except for an optional concise pointer when useful.
|
|
81
81
|
|
|
82
|
-
For follow-up entries, use `journal` or `decision` and link both the prior entry
|
|
82
|
+
For follow-up entries, use `journal-entry` or `journal-decision` and link both the prior entry
|
|
83
83
|
and recalled evidence.
|
|
84
84
|
|
|
85
85
|
After every write, re-read changed files and link targets. Verify IDs, paths,
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: research-codebase
|
|
2
|
+
name: journal-research-codebase
|
|
3
3
|
description: Research and document how a topic currently works in the codebase, producing a canonical supporting entry. Use for implementation maps, integration analysis, type discovery, usage patterns, tests, and dependencies before planning or changing code.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: research-web
|
|
2
|
+
name: journal-research-web
|
|
3
3
|
description: Research a current external topic on the internet and write a sourced canonical supporting entry. Use for official documentation, standards, libraries, best practices, examples, or other temporally unstable external knowledge relevant to active work.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -1,12 +1,17 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: resume
|
|
3
|
-
description:
|
|
2
|
+
name: journal-resume
|
|
3
|
+
description: Restore djournal work-item context from canonical Markdown when continuing a project, asking where we left off or what is next, switching work, or recovering after compaction. Read-only project continuity; distinct from reopening a chat session or loading native agent memory.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
6
|
# Resume active work
|
|
7
7
|
|
|
8
8
|
Operate read-only.
|
|
9
9
|
|
|
10
|
+
Read the selected work item's canonical state and latest entry before reporting
|
|
11
|
+
continuity. A resumed conversation or native memory summary can provide hints,
|
|
12
|
+
but does not establish the journal's current state. Reuse canonical context
|
|
13
|
+
already loaded for this task when it is still current.
|
|
14
|
+
|
|
10
15
|
## Load
|
|
11
16
|
|
|
12
17
|
1. Read completely:
|
|
@@ -23,7 +28,7 @@ Operate read-only.
|
|
|
23
28
|
is available, ask the user to choose one before doing meaningful writes.
|
|
24
29
|
After the user chooses, bind the session with
|
|
25
30
|
`djournal work bind <slug> --session <id>` when a session id is available,
|
|
26
|
-
or use `switch` only when the user wants to change the global fallback.
|
|
31
|
+
or use `journal-switch` only when the user wants to change the global fallback.
|
|
27
32
|
3. If `work.md` is legacy/missing, infer a temporary title/slug and clearly mark
|
|
28
33
|
metadata as incomplete. Do not repair it.
|
|
29
34
|
4. Index entry frontmatter under `journal/`, `_research/`, `docs/`, and
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
name: switch
|
|
2
|
+
name: journal-switch
|
|
3
3
|
description: Change the active filesystem journal work item without changing Git state. Use when moving journal context to another existing project or work stream.
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -28,10 +28,10 @@ not Git branches. For parallel sessions, prefer binding the session with
|
|
|
28
28
|
using the exact shape in `STATE.md`.
|
|
29
29
|
9. Re-read state and verify the target folder and `work.md` slug.
|
|
30
30
|
10. Present the selected work's latest spine summary and next steps using the
|
|
31
|
-
`resume` retrieval procedure.
|
|
31
|
+
`journal-resume` retrieval procedure.
|
|
32
32
|
|
|
33
33
|
## Constraints
|
|
34
34
|
|
|
35
35
|
- Do not modify work-item status, visibility, IDs, or timestamps.
|
|
36
|
-
- Do not create a missing work item; use `init-work` instead.
|
|
36
|
+
- Do not create a missing work item; use `journal-init-work` instead.
|
|
37
37
|
- Never infer an ambiguous write target from recency.
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: journal-workflow
|
|
3
|
-
description: Coordinate
|
|
3
|
+
description: Coordinate djournal project memory for meaningful implementation, investigation, decisions, and status changes. Select and load the required journal-* skills for context, evidence, planning, and closure even when the user does not mention journaling.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
6
|
# Coordinate the journal workflow
|
|
@@ -14,10 +14,29 @@ the underlying journal skills.
|
|
|
14
14
|
2. Read `.agents/rules/STATE.md` before any write.
|
|
15
15
|
3. Load other journal rules only through the selected underlying skill.
|
|
16
16
|
|
|
17
|
+
## Select and load skills
|
|
18
|
+
|
|
19
|
+
Follow the intent routes and invocation contract in `AUTOMATION.md`. Invoke the
|
|
20
|
+
exact `journal-*` skill through the harness skill tool when available; otherwise
|
|
21
|
+
read `.agents/skills/<skill-name>/SKILL.md`. Do this before the selected work,
|
|
22
|
+
not merely by mentioning the skill or doing a CLI search.
|
|
23
|
+
|
|
24
|
+
- Continue or recover active work: `journal-resume` when journal context is
|
|
25
|
+
needed and is not already loaded.
|
|
26
|
+
- Answer from past project decisions or history: `journal-recall` and its
|
|
27
|
+
canonical evidence, even when native memory suggests an answer.
|
|
28
|
+
- Plan a multi-phase or risky change: `journal-plan` before implementation,
|
|
29
|
+
unless an applicable accepted plan already exists.
|
|
30
|
+
|
|
31
|
+
Native conversation resume, auto memory, and plan mode remain useful but do not
|
|
32
|
+
complete these journal steps. Preserve read-only restrictions and the existing
|
|
33
|
+
opt-out, trivial-work, and plan-threshold rules. No repeated invocation is needed
|
|
34
|
+
for unchanged skill instructions already loaded in this context.
|
|
35
|
+
|
|
17
36
|
## Classify
|
|
18
37
|
|
|
19
38
|
- Honor `journal: off` and finish with `<!-- journal-status: off -->`.
|
|
20
|
-
- For read-only work, use `recall` or `resume` only when history matters. Do not
|
|
39
|
+
- For read-only work, use `journal-recall` or `journal-resume` only when history matters. Do not
|
|
21
40
|
write an entry; finish with `<!-- journal-status: not-needed -->` unless
|
|
22
41
|
recalled evidence materially affects current or closed work as described
|
|
23
42
|
below.
|
|
@@ -27,16 +46,16 @@ the underlying journal skills.
|
|
|
27
46
|
|
|
28
47
|
## Recall reinforcement
|
|
29
48
|
|
|
30
|
-
Keep `recall` itself read-only. After a recall/resume answer, decide whether the
|
|
49
|
+
Keep `journal-recall` itself read-only. After a recall/resume answer, decide whether the
|
|
31
50
|
retrieved evidence materially affects current journal work before finalizing:
|
|
32
51
|
|
|
33
52
|
- Immaterial recall: answer from sources and finish `not-needed`.
|
|
34
|
-
- Current closure in progress: pass the recalled source entries to `journal` so
|
|
53
|
+
- Current closure in progress: pass the recalled source entries to `journal-entry` so
|
|
35
54
|
the new spine entry links them with `references`, `relates_to`, or
|
|
36
55
|
`supersedes` as appropriate.
|
|
37
|
-
- Explicit current/closed target: use `reinforce` to add conservative links or a
|
|
56
|
+
- Explicit current/closed target: use `journal-reinforce` to add conservative links or a
|
|
38
57
|
follow-up spine entry.
|
|
39
|
-
- Accepted contradiction or replacement: use `decision` or `journal` to record
|
|
58
|
+
- Accepted contradiction or replacement: use `journal-decision` or `journal-entry` to record
|
|
40
59
|
the correction and link the recalled evidence.
|
|
41
60
|
|
|
42
61
|
Material evidence includes recalled sources that confirm a claim being recorded,
|
|
@@ -64,15 +83,15 @@ Pass materially used entries to the closing journal skill for typed citations.
|
|
|
64
83
|
|
|
65
84
|
## Run checkpoints
|
|
66
85
|
|
|
67
|
-
1. **Start:** use `resume` when active-work history can affect the task. Use
|
|
68
|
-
`init-work` only for an unambiguous new durable work stream.
|
|
69
|
-
2. **Plan:** use `plan` for multi-phase, risky, cross-component, or
|
|
86
|
+
1. **Start:** use `journal-resume` when active-work history can affect the task. Use
|
|
87
|
+
`journal-init-work` only for an unambiguous new durable work stream.
|
|
88
|
+
2. **Plan:** use `journal-plan` for multi-phase, risky, cross-component, or
|
|
70
89
|
decision-heavy implementation. Do not plan again when an accepted plan exists.
|
|
71
|
-
3. **Support:** route durable evidence to `research-codebase`, `research-web`,
|
|
72
|
-
`decision`, `document`, or `reinforce` as appropriate.
|
|
90
|
+
3. **Support:** route durable evidence to `journal-research-codebase`, `journal-research-web`,
|
|
91
|
+
`journal-decision`, `journal-document`, or `journal-reinforce` as appropriate.
|
|
73
92
|
4. **Implement and validate:** make the requested changes and gather bounded,
|
|
74
93
|
deterministic evidence.
|
|
75
|
-
5. **Close:** check the latest spine entry for duplication, then use `journal`
|
|
94
|
+
5. **Close:** check the latest spine entry for duplication, then use `journal-entry`
|
|
76
95
|
once when meaningful state changed.
|
|
77
96
|
|
|
78
97
|
## Finish
|
package/AGENTS.md
CHANGED
|
@@ -9,12 +9,21 @@ For every request:
|
|
|
9
9
|
|
|
10
10
|
1. Honor an explicit `journal: off` instruction for that request.
|
|
11
11
|
2. Classify the work using `.agents/rules/AUTOMATION.md`.
|
|
12
|
-
3.
|
|
13
|
-
|
|
12
|
+
3. Load `journal-workflow` for meaningful mutations, durable research,
|
|
13
|
+
decisions, documentation, or material status changes. Invoke the exact skill
|
|
14
|
+
name through the harness skill tool when available; otherwise read
|
|
15
|
+
`.agents/skills/journal-workflow/SKILL.md` and follow it.
|
|
14
16
|
4. Do not force a plan or journal entry for read-only or trivial work.
|
|
15
17
|
5. Before the final response, close meaningful changed state through the
|
|
16
18
|
existing journal skills. Hooks never write entries.
|
|
17
19
|
|
|
20
|
+
Journal skills use the `journal-*` namespace. Route project continuity to
|
|
21
|
+
`journal-resume`, past decisions/history to `journal-recall`, and work meeting
|
|
22
|
+
the planning threshold to `journal-plan`. Load the selected skill before its
|
|
23
|
+
work; native conversation resume, auto memory, and plan mode do not perform
|
|
24
|
+
these journal procedures. Use native tools alongside the journal as useful.
|
|
25
|
+
After compaction or a work switch, reload journal context only as needed.
|
|
26
|
+
|
|
18
27
|
End every final response with exactly one hidden status marker:
|
|
19
28
|
|
|
20
29
|
- `<!-- journal-status: closed .journal/work/<slug>/journal/<entry>.md -->`
|
package/README.md
CHANGED
|
@@ -275,7 +275,7 @@ djournal recall search "billing webhook retries" --type decision,research --limi
|
|
|
275
275
|
```
|
|
276
276
|
|
|
277
277
|
Or install a harness and ask its agent naturally: “Recall why we chose the
|
|
278
|
-
current billing webhook retry strategy.” The bundled `recall` skill calls the
|
|
278
|
+
current billing webhook retry strategy.” The bundled `journal-recall` skill calls the
|
|
279
279
|
same CLI, selects the best journal paths, reads the canonical Markdown, follows
|
|
280
280
|
relevant links, and returns an evidence-backed answer.
|
|
281
281
|
|
|
@@ -284,6 +284,10 @@ items and entry paths before reading canonical Markdown. Warm queries reuse the
|
|
|
284
284
|
snapshot; additions, edits, deletes, and work-metadata changes refresh it on the
|
|
285
285
|
next search. Use `--no-cache` for a one-call in-memory index. Cache corruption or
|
|
286
286
|
write denial is recoverable and never triggers a share, sync, commit, or push.
|
|
287
|
+
All bundled skills use the `journal-*` namespace, including `journal-resume`,
|
|
288
|
+
`journal-plan`, and `journal-entry`. See [skill names and invocation](docs/installation.md#journal-skill-names-and-invocation)
|
|
289
|
+
for the rename mapping and upgrade behavior.
|
|
290
|
+
|
|
287
291
|
See [Fast recall](docs/recall.md) for standalone CLI examples, agent behavior,
|
|
288
292
|
filters, and cache lifecycle details.
|
|
289
293
|
|
package/docs/installation.md
CHANGED
|
@@ -283,6 +283,54 @@ Claude integration removes its unchanged installed skills while retaining
|
|
|
283
283
|
`.agents/skills/`. Existing or locally modified skill files are preserved;
|
|
284
284
|
different content at an installation destination produces an `ASSET_CONFLICT`.
|
|
285
285
|
|
|
286
|
+
### Journal skill names and invocation
|
|
287
|
+
|
|
288
|
+
All bundled skills use `journal-*` names. Their folder and frontmatter names
|
|
289
|
+
match across `.agents/skills/`, Claude's project copies, and Grok registration.
|
|
290
|
+
|
|
291
|
+
| Previous skill | Current skill |
|
|
292
|
+
| --- | --- |
|
|
293
|
+
| `resume` | `journal-resume` |
|
|
294
|
+
| `recall` | `journal-recall` |
|
|
295
|
+
| `plan` | `journal-plan` |
|
|
296
|
+
| `journal` | `journal-entry` |
|
|
297
|
+
| `init-work` | `journal-init-work` |
|
|
298
|
+
| `switch` | `journal-switch` |
|
|
299
|
+
| `decision` | `journal-decision` |
|
|
300
|
+
| `document` | `journal-document` |
|
|
301
|
+
| `research-codebase` | `journal-research-codebase` |
|
|
302
|
+
| `research-web` | `journal-research-web` |
|
|
303
|
+
| `reinforce` | `journal-reinforce` |
|
|
304
|
+
| `audit` | `journal-audit` |
|
|
305
|
+
| `reconcile` | `journal-reconcile` |
|
|
306
|
+
| `journal-workflow` | `journal-workflow` (unchanged) |
|
|
307
|
+
|
|
308
|
+
In Claude, invoke `/journal-resume`, `/journal-recall`, or `/journal-plan`
|
|
309
|
+
directly. In Codex, use `$journal-resume`, `$journal-recall`, or `$journal-plan`.
|
|
310
|
+
Natural requests such as “where did we leave off?” or “why did we choose this?”
|
|
311
|
+
should route through the corresponding skill when project memory is needed.
|
|
312
|
+
If a skill is missing from the catalog, the workflow reads its canonical
|
|
313
|
+
`.agents/skills/<skill-name>/SKILL.md` file. If that file is also missing,
|
|
314
|
+
upgrade or repair the installation; do not assume the journal was consulted.
|
|
315
|
+
|
|
316
|
+
Run `djournal upgrade` after updating the CLI to migrate managed copies.
|
|
317
|
+
Unchanged installer-created old skill files are removed. Modified managed
|
|
318
|
+
files produce a conflict and remain intact; pre-existing unowned copies are
|
|
319
|
+
preserved. No generic aliases are installed. Manually installed personal skills
|
|
320
|
+
are outside this migration: review those separately if old names remain visible.
|
|
321
|
+
Restart the agent after upgrading, and re-register Grok workflows as needed.
|
|
322
|
+
CLI commands such as `djournal recall` and entry types such as `plan` keep
|
|
323
|
+
their names.
|
|
324
|
+
|
|
325
|
+
Claude's native `/resume` and `/plan` commands and auto memory serve different
|
|
326
|
+
purposes from journal context retrieval and durable planning. The namespace and
|
|
327
|
+
explicit routing help distinguish them; native tools remain available. Automatic
|
|
328
|
+
selection still depends on the model, available skills, permissions, and context
|
|
329
|
+
budget. These instructions do not guarantee a skill call. If selection is skipped,
|
|
330
|
+
invoke the journal skill directly and confirm it cites or writes the expected
|
|
331
|
+
canonical journal entry. See [Claude's skill selection and troubleshooting](https://code.claude.com/docs/en/skills)
|
|
332
|
+
and [native commands](https://code.claude.com/docs/en/commands).
|
|
333
|
+
|
|
286
334
|
### Missing instruction blocks
|
|
287
335
|
|
|
288
336
|
Reinstall and upgrade restore a completely missing djournal block in `AGENTS.md`
|
|
@@ -15,8 +15,8 @@ read-only setup is achieved by combining three controls:
|
|
|
15
15
|
push to it.
|
|
16
16
|
- djournal sync config that enables `pull` but does not enable automatic
|
|
17
17
|
publication.
|
|
18
|
-
- Agent prompts that ask for read-only skills such as `recall
|
|
19
|
-
`
|
|
18
|
+
- Agent prompts that ask for read-only skills such as `journal-recall` and
|
|
19
|
+
`journal-resume`, not implementation or journal closure.
|
|
20
20
|
|
|
21
21
|
This still writes local files under the user's djournal home. "Read-only" means
|
|
22
22
|
read-only with respect to the organization remote, not a filesystem mode that
|
package/docs/recall.md
CHANGED
|
@@ -2,7 +2,12 @@
|
|
|
2
2
|
|
|
3
3
|
Recall finds the journal files most likely to answer a question before anyone
|
|
4
4
|
spends time opening and summarizing them. It can be used directly as a small
|
|
5
|
-
search CLI, or indirectly through an agent that has djournal's `recall` skill.
|
|
5
|
+
search CLI, or indirectly through an agent that has djournal's `journal-recall` skill.
|
|
6
|
+
|
|
7
|
+
For an explicit skill invocation, use `/journal-recall` in Claude or
|
|
8
|
+
`$journal-recall` in Codex. Ask `journal-resume` for the selected work item's
|
|
9
|
+
current state and next steps. Native agent memory may help locate information,
|
|
10
|
+
but journal answers cite the canonical Markdown sources.
|
|
6
11
|
Both paths use the same local index and keep Markdown as the source of truth.
|
|
7
12
|
|
|
8
13
|
## Use it as a standalone CLI
|
|
@@ -65,7 +70,7 @@ Then ask naturally:
|
|
|
65
70
|
Recall why we chose the current billing webhook retry strategy.
|
|
66
71
|
```
|
|
67
72
|
|
|
68
|
-
The `recall` skill classifies the question, calls the same CLI with `--json`,
|
|
73
|
+
The `journal-recall` skill classifies the question, calls the same CLI with `--json`,
|
|
69
74
|
uses the ranked paths to select a small evidence set, and reads those canonical
|
|
70
75
|
Markdown files before answering. It follows relevant entry links and starts
|
|
71
76
|
from the latest spine entry when the question needs chronology or current
|
package/docs/remote-sync.md
CHANGED
|
@@ -146,7 +146,7 @@ response that closes `.journal/work/<slug>/journal/...` synchronizes that
|
|
|
146
146
|
`<slug>` even if `state.json` currently selects a different active work item.
|
|
147
147
|
|
|
148
148
|
Configured standalone automatic sync also pulls when a new chat/session starts.
|
|
149
|
-
The `init-work` and `switch` workflows pull before enumerating work, and an
|
|
149
|
+
The `journal-init-work` and `journal-switch` workflows pull before enumerating work, and an
|
|
150
150
|
explicit request such as “sync the journal” triggers a pull before the agent
|
|
151
151
|
handles publication. Automatic hook failures are reported as bounded context;
|
|
152
152
|
explicit pull/sync conflicts exit nonzero for manual resolution.
|
|
@@ -4,28 +4,28 @@ Grok Bot loads shared workflows. It does not load `.agents/skills` by itself
|
|
|
4
4
|
and it has no session hooks. After `djournal install --harness grok-bot`,
|
|
5
5
|
register each installed skill with the harness **UpdateSkill** action.
|
|
6
6
|
|
|
7
|
-
Keep the upstream skill files
|
|
7
|
+
Keep the upstream skill files portable. Rewrite a copy for the payload. Do not
|
|
8
8
|
edit the installed `SKILL.md` files in place, and do not embed a workspace
|
|
9
9
|
path, owner name, or agent id.
|
|
10
10
|
|
|
11
|
-
##
|
|
11
|
+
## Workflow ids
|
|
12
12
|
|
|
13
|
-
| Installed skill directory |
|
|
13
|
+
| Installed skill directory | Workflow id |
|
|
14
14
|
| --- | --- |
|
|
15
15
|
| `journal-workflow` | `journal-workflow` |
|
|
16
|
-
| `resume` | `journal-resume` |
|
|
17
|
-
| `init-work` | `journal-init-work` |
|
|
18
|
-
| `switch` | `journal-switch` |
|
|
19
|
-
| `journal` | `journal-entry` |
|
|
20
|
-
| `plan` | `journal-plan` |
|
|
21
|
-
| `recall` | `journal-recall` |
|
|
22
|
-
| `decision` | `journal-decision` |
|
|
23
|
-
| `document` | `journal-document` |
|
|
24
|
-
| `research-codebase` | `journal-research-codebase` |
|
|
25
|
-
| `research-web` | `journal-research-web` |
|
|
26
|
-
| `reinforce` | `journal-reinforce` |
|
|
27
|
-
| `audit` | `journal-audit` |
|
|
28
|
-
| `reconcile` | `journal-reconcile` |
|
|
16
|
+
| `journal-resume` | `journal-resume` |
|
|
17
|
+
| `journal-init-work` | `journal-init-work` |
|
|
18
|
+
| `journal-switch` | `journal-switch` |
|
|
19
|
+
| `journal-entry` | `journal-entry` |
|
|
20
|
+
| `journal-plan` | `journal-plan` |
|
|
21
|
+
| `journal-recall` | `journal-recall` |
|
|
22
|
+
| `journal-decision` | `journal-decision` |
|
|
23
|
+
| `journal-document` | `journal-document` |
|
|
24
|
+
| `journal-research-codebase` | `journal-research-codebase` |
|
|
25
|
+
| `journal-research-web` | `journal-research-web` |
|
|
26
|
+
| `journal-reinforce` | `journal-reinforce` |
|
|
27
|
+
| `journal-audit` | `journal-audit` |
|
|
28
|
+
| `journal-reconcile` | `journal-reconcile` |
|
|
29
29
|
|
|
30
30
|
Hub-only prose in this recipe is not an upstream skill:
|
|
31
31
|
|
|
@@ -39,7 +39,7 @@ Hub-only prose in this recipe is not an upstream skill:
|
|
|
39
39
|
1. Set `<project-root>` to the install target (the directory that contains
|
|
40
40
|
`.djournal.json` and `.agents/`).
|
|
41
41
|
2. Read `<project-root>/.agents/skills/<directory>/SKILL.md`.
|
|
42
|
-
3. Take `name` and `description` from the YAML frontmatter. Use the
|
|
42
|
+
3. Take `name` and `description` from the YAML frontmatter. Use the same namespaced
|
|
43
43
|
workflow id from the table. Do not invent a project-specific id.
|
|
44
44
|
4. Rewrite the body into the UpdateSkill payload:
|
|
45
45
|
- Point rule reads at `<project-root>/.agents/rules/...`. Journal root
|
package/lib/installer/index.js
CHANGED
|
@@ -601,7 +601,7 @@ function readActiveWork(context, requested) {
|
|
|
601
601
|
slug = state.active_work_name;
|
|
602
602
|
}
|
|
603
603
|
if (typeof slug !== "string" || !slug) {
|
|
604
|
-
throw new InstallerError("No active work found. Run init-work or switch first.", "NO_ACTIVE_WORK");
|
|
604
|
+
throw new InstallerError("No active work found. Run journal-init-work or journal-switch first.", "NO_ACTIVE_WORK");
|
|
605
605
|
}
|
|
606
606
|
if (slug.includes("/") || slug.includes("\\") || slug === "." || slug === "..") {
|
|
607
607
|
throw new InstallerError(`invalid work slug: ${slug}`, "INVALID_WORK");
|
package/lib/setup.js
CHANGED
|
@@ -209,7 +209,7 @@ async function setup(options, dependencies = {}) {
|
|
|
209
209
|
try { result = await api.listWork(base); }
|
|
210
210
|
catch (error) {
|
|
211
211
|
if (error.code !== "NO_ACTIVE_WORK") throw error;
|
|
212
|
-
ui.write("No work items yet. Ask your coding agent to start a new work item with init-work.");
|
|
212
|
+
ui.write("No work items yet. Ask your coding agent to start a new work item with journal-init-work.");
|
|
213
213
|
return;
|
|
214
214
|
}
|
|
215
215
|
const selected = await menu("Choose a work item", [
|
package/package.json
CHANGED
package/spec.md
CHANGED
|
@@ -42,18 +42,20 @@ project/
|
|
|
42
42
|
│ │ ├── SAFETY.md
|
|
43
43
|
│ │ └── STATE.md
|
|
44
44
|
│ └── skills/
|
|
45
|
-
│ ├── audit/
|
|
46
|
-
│ ├── decision/
|
|
47
|
-
│ ├── document/
|
|
48
|
-
│ ├── init-work/
|
|
49
|
-
│ ├── journal/
|
|
50
|
-
│ ├──
|
|
51
|
-
│ ├──
|
|
52
|
-
│ ├──
|
|
53
|
-
│ ├──
|
|
54
|
-
│ ├──
|
|
55
|
-
│ ├──
|
|
56
|
-
│
|
|
45
|
+
│ ├── journal-audit/
|
|
46
|
+
│ ├── journal-decision/
|
|
47
|
+
│ ├── journal-document/
|
|
48
|
+
│ ├── journal-init-work/
|
|
49
|
+
│ ├── journal-entry/
|
|
50
|
+
│ ├── journal-workflow/
|
|
51
|
+
│ ├── journal-reinforce/
|
|
52
|
+
│ ├── journal-plan/
|
|
53
|
+
│ ├── journal-recall/
|
|
54
|
+
│ ├── journal-reconcile/
|
|
55
|
+
│ ├── journal-research-codebase/
|
|
56
|
+
│ ├── journal-research-web/
|
|
57
|
+
│ ├── journal-resume/
|
|
58
|
+
│ └── journal-switch/
|
|
57
59
|
└── .journal/
|
|
58
60
|
├── state.json
|
|
59
61
|
└── work/
|
|
@@ -209,15 +211,15 @@ original wall-clock creation time.
|
|
|
209
211
|
Typical flow:
|
|
210
212
|
|
|
211
213
|
```text
|
|
212
|
-
init-work or switch
|
|
213
|
-
→ resume
|
|
214
|
-
→ research-codebase / research-web
|
|
215
|
-
→ plan
|
|
214
|
+
journal-init-work or journal-switch
|
|
215
|
+
→ journal-resume
|
|
216
|
+
→ journal-research-codebase / journal-research-web
|
|
217
|
+
→ journal-plan
|
|
216
218
|
→ implementation
|
|
217
|
-
→ decision / document when needed
|
|
218
|
-
→ journal
|
|
219
|
-
→ audit or reconcile periodically
|
|
220
|
-
→ recall for historical questions
|
|
219
|
+
→ journal-decision / journal-document when needed
|
|
220
|
+
→ journal-entry
|
|
221
|
+
→ journal-audit or journal-reconcile periodically
|
|
222
|
+
→ journal-recall for historical questions
|
|
221
223
|
```
|
|
222
224
|
|
|
223
225
|
Producer skills emit exact metadata and follow the shared safety/link contracts.
|
|
@@ -225,12 +227,12 @@ Retrieval skills index frontmatter first, then deep-read only selected bodies.
|
|
|
225
227
|
|
|
226
228
|
## 12. Resume and recall
|
|
227
229
|
|
|
228
|
-
`resume` loads work metadata, latest spine state/timeline/next steps, five recent
|
|
230
|
+
`journal-resume` loads work metadata, latest spine state/timeline/next steps, five recent
|
|
229
231
|
spine entries as fallback, and at most three relevant linked supporting entries.
|
|
230
232
|
It reports possible unjournaled Git changes and unmatched plans without claiming
|
|
231
233
|
formal session recovery.
|
|
232
234
|
|
|
233
|
-
`recall` ranks work items from `work.md`, filters entries by metadata, and follows
|
|
235
|
+
`journal-recall` ranks work items from `work.md`, filters entries by metadata, and follows
|
|
234
236
|
typed links lazily. Deep reads are bounded to eight spine and three supporting
|
|
235
237
|
entries; technical verification is bounded to five code files.
|
|
236
238
|
|
|
@@ -239,7 +241,7 @@ filename, and heading.
|
|
|
239
241
|
|
|
240
242
|
## 13. Audit and reconciliation
|
|
241
243
|
|
|
242
|
-
`audit` is strictly read-only and walks every entry in scope. It checks the
|
|
244
|
+
`journal-audit` is strictly read-only and walks every entry in scope. It checks the
|
|
243
245
|
filesystem contract plus the reconciliation checklist:
|
|
244
246
|
|
|
245
247
|
1. misclassified type
|
|
@@ -250,7 +252,7 @@ filesystem contract plus the reconciliation checklist:
|
|
|
250
252
|
6. wrong work status
|
|
251
253
|
7. forked/duplicate entry numbers, flag only
|
|
252
254
|
|
|
253
|
-
`reconcile` runs audit first, reports proposed changes, then applies only clear
|
|
255
|
+
`journal-reconcile` runs audit first, reports proposed changes, then applies only clear
|
|
254
256
|
fixes. It may enrich legacy metadata, repair deterministic links, improve factual
|
|
255
257
|
summaries, extract explicit decisions, and correct unequivocal status.
|
|
256
258
|
|