Memory Sync
This is a Hermes-native memory-sync workflow skill.
Why This Exists
memory-sync exists so Hermes users can ask for this workflow in chat and receive a structured, evidence-bounded OMH operating surface instead of ad hoc narration.
Do Not Use When
- The request is already handled by a narrower explicit skill with stronger evidence.
- The user asks OMH to secretly run external platforms, connectors, schedulers, file exports, or runtime agents.
- The only safe answer is to ask for missing authority, credentials, target, or observed evidence first.
Examples
Good example:
- Prompt: memory-sync inspect stale MEMORY.md claims, prepare a native write diff, and ask which claims to keep, revise, or archive.
- Expected behavior: Produce
prepare_memory_sync with required context, wrapper actions, and not-evidence boundaries.
- Why: The prompt names a real workflow surface that Hermes can orchestrate without hiding execution.
Bad example:
- Prompt: memory-sync claim a prepared native diff changed MEMORY.md or USER.md.
- Expected behavior: Report the missing observed evidence or authority instead of claiming the external step happened.
- Why: Prepared OMH guidance is not platform, runtime, connector, file, memory, or delivery evidence.
Completion Checklist
- Confirm the workflow target, evidence boundary, and stop condition are named.
- Report which outputs are prepared, observed, blocked, or missing.
- Name the smallest next verification or handoff instead of claiming completion from narration.
Recovery Notes
- If required context is missing, ask one blocking question or route back to the narrower workflow.
- If runtime or wrapper evidence is unavailable, keep the status as not_observed and expose the next observable action.
Workflow Lane
- Current lane: Retained knowledge (
memory-new, memory-sync, decision-recall, wiki) - memory, rejected alternatives, wiki notes, retrieval, and staleness.
- If intent belongs to another lane, hand back to
oh-my-hermes or name the adjacent workflow.
- Shared product, routing, compatibility, and evidence rules:
omh-routing/references/skill-common-rail.md.
English-Canonical Interview Protocol
- Inventory (목록) - Nobody hands you the material: call
omh_memory with action="status" to get the entry inventory - per-file entry counts, per-entry index and size, headroom, and which entries have no OMH record. It returns counts and hashes, never entry text, because the text is already yours.
- Claim extraction (추출) - Break the
USER.md and MEMORY.md material into claims. Quote only observed claims; never invent provenance.
- Provenance (출처) - Ask for the source class and distinguish Hermes-native, provider, and vector material as
not_omh_reviewed.
- Target (대상) - Review existing native-memory claims only. Route a new project/product fact to
memory-new.
- Candidate selection (후보) - Default to a short interview, not a census: pick about five candidate entries per pass and say why each was picked. Rank by the signals already on hand — dreaming reminders (duplicate clusters, deadline,
stale_review_required), the status bridge's OMH-record similarity rows (near-duplicates), and claims that look stale, conflicting, or overgeneralized. Walk the complete inventory only when the user asks for a full review.
- Per-entry confirmation (확인) - Walk the selected candidates one at a time. For each entry, quote it back from your own memory file and state what you take it to mean, then ask the user to keep, revise, or archive it before moving on. Do not summarize the whole file and ask one question about all of it; a review the user cannot correct entry by entry is not a review.
- Cursor (이어하기) - Close every pass by naming what was covered and what was not — reviewed entry indexes, remaining candidates, and the next entry a resumed review would start from — so an interrupted interview resumes instead of restarting. The resume point lives only in the conversation; no store persists it, so name it explicitly rather than assuming the system remembers.
- Review (검토) - Prioritize stale, conflicting, duplicate, and overgeneralized claims. Offer keep, revise, or archive choices; do not describe an archive as removal.
- Attention (주의) - For a reviewed OMH-local record, keep/archive is an attention tier:
active leads the working context, reference stays recallable behind active peers, archive leaves default recall. Preview with omh memory attention <record-id> --tier <tier>, say which records stay in the working context and which leave it, then apply with --apply only after the user agrees. The preview writes nothing.
- Diff (차이) - Prepare one concise native write diff with before/after claims and counts. Keep the caps: MEMORY.md about 2,200 characters and USER.md about 1,375 characters.
- Native-write boundary (쓰기) - OMH prepares guidance and a native write diff only; no OMH surface invokes, applies, or observes a
MEMORY.md/USER.md write.
- Apply after approval (적용) - The interview does not end at a diff. Per-entry keep/revise/archive answers are input to the diff, not approval of it: ask for one explicit approval of the assembled diff. After that approval, apply the approved entries yourself through the Hermes-native memory tool — the same tool that owns these files — and report what the write observably changed; with the tool available, leaving an approved diff silently unapplied fails the interview. If the native memory tool is unavailable, report the approved diff and stop: that is a completed review with the write pending, never a failed interview and never a reason to edit the files directly. The OMH artifact stays
memory_curation_review/v1 metadata either way: the native write is Hermes's own act and never becomes OMH mutation evidence.
Memory Boundaries
The prepared artifact is memory_curation_review/v1, not native-memory mutation evidence. Hermes-native and external provider/vector context is not_omh_reviewed: it can nominate an OMH candidate but never inherits OMH approval. A configured Hermes runtime may transmit rendered OMH prefetch content in its model request.
Use lifecycle words literally: expire removes influence only; retire archives recoverably; restore creates a new pending revision while preserving the archive; prune hard-deletes only the manifest-declared OMH-local target set. Report restore and prune first. No lifecycle result proves anything outside that named local target set.
An attention tier is not a lifecycle state, and the two uses of "archive" are different: the archive tier only stops a record from occupying the default working context, leaving it in the store, readable, and answerable by omh memory recall --include-archived, while retire moves an expired revision into the local archive directory. Neither is deletion; never describe either as one.
Legacy v1 material is migration/review-required. Present memory inventory counts and the report-first per-artifact memory reactivate ... --apply path; inventory and reactivation never silently grant replay eligibility.
Dreaming runs automatically in reminder mode at five scheduler points: turn when the interval is due (default five turns), compaction before compression discards messages, session_end after a productive session, shutdown as the final process opportunity, and session_start_recovery when the prior session ended without consolidation. It prepares reminders for duplicate clusters, records at or near their deadline, headroom below the configured floor, stale_review_required, and expired_volatile_records; an unchanged standing condition is suppressed until its value changes. Anything whose source OMH cannot explain is not a candidate. Dreaming never invokes a model or performs consolidation, retirement, restore, or prune.
Treat ranking signals within their limits: pins guarantee inclusion but never override expiry, scope, perspective, or review eligibility; attention tiers control working-context occupancy, not truth; approved_manual has 100% veracity weight and approved_auto_safe 90%, while an unknown approval mode fails closed to the lower weight; age only breaks ties within an equal relevance rank; and usage uses saturating buckets so repeated delivery cannot compound into a permanent lead.
Normal users use natural-language Hermes chat. omh memory ... commands are agent/operator control-plane references, not normal-user setup.
Use When
Use when existing Hermes USER.md, MEMORY.md, or accumulated skill memories need an English-canonical, claim-by-claim review. It prepares native write guidance only; no OMH surface invokes, applies, or observes a native write — a user-approved diff is applied by Hermes's own native memory tool. Do not use for new project or product candidates.
Strong routing signals: `memory-sync`, `memory curation`, `memory review`, `memory inspect`, `memory check`, `memory update`, `context cleanup`, `curate memory`, `stale memory`, `hermes remembers`, `conflicting memory`, `duplicate skill`, `MEMORY.md`, `USER.md`, `what you remember about me`, `your memory about me`, `your memories`, `memory interview`, `memories still true`, `기억하고 있는`, `기억하고 있는 프로젝트 맥락`, `기억하는 맥락`, `현재 hermes가 기억하는 맥락`, `현재 헤르메스가 기억하는 맥락`, `헤르메스가 기억하는 맥락`, `오래된 맥락`, `오래된 기억`, `기억 점검`, `기억 정리`, `메모리 업데이트`, `메모리 검사`, `메모리 점검`, `메모리 정리`, `맥락 점검`, `맥락 정리`, `맥락 피드백`, `등록된 맥락`, `헤르메스 기억`, `중복 스킬`, `나에 대해 잘못 알고`, `저장된 내 정보`, `너한테 저장된`, `저장된 프로필`, `기억 바로잡`, `메모리 인터뷰`, `기억 인터뷰`
Catalog Metadata
Category: memory
Phase: curation-review
Hermes role: memory-keeper
Quality tier: workflow-surface-gated
Reasoning demand: light
Quality bar:
- Name the user-facing workflow objective, required context, next action, and stop condition.
- Separate prepared guidance from observed platform, runtime, connector, file, memory, or delivery evidence.
- Expose missing tools, credentials, targets, or observations as user-visible gaps.
- State that Hermes-native and external provider/vector context is not_omh_reviewed, can nominate a candidate only, and may receive rendered OMH prefetch content through a configured Hermes runtime model request.
Handoff policy:
Keep this as Hermes-facing orchestration guidance first. Prepare executor, connector, gateway, or host-runtime handoff only when the user accepts that next step and observed evidence can be recorded.
Required inputs:
- user request
- target context
- delivery or status expectation
- known missing evidence
Expected outputs:
- memory-sync/v1 card or guidance
- next action
- prepared-vs-observed boundary
Artifact expectations:
- memory-sync/v1 metadata-only runtime or wrapper card when recorded
Safety rules:
- A memory-sync review is prompt guidance only: no OMH surface invokes, applies, or observes a MEMORY.md or USER.md write. Applying a user-approved diff is Hermes's own act through its native memory tool — report it as an observed native write, never as OMH mutation evidence. Hermes-native and external provider/vector context is not_omh_reviewed and never inherits OMH approval.
- Do not claim connector, gateway, runtime, file generation, memory mutation, or host automation evidence from prepared guidance.
- Keep English as the canonical protocol; Korean routing triggers and concise Korean help labels remain available.
- Quote claims only when observed, do not invent provenance, and keep the prepared native diff separate from any native write.
Runtime Evidence
Preferred harness for this skill: memory-sync.
omh runtime record --skill memory-sync --harness memory-sync --status started
Record observed delegation results; otherwise return not_available or not_observed.
Prepared OMH routing is not execution, review, CI, merge-readiness, or merge evidence.
- Treat wrapper memory/context summaries as advisory local context, not proof of opaque Hermes memory reads or changes.
Preserve workflow intent and stop conditions; verify before claiming completion.
Use Hermes-native subagent/delegation features when available: native subagents -> Hermes delegation when available, otherwise sequential lanes.
Shared product, compatibility, topology, memory, harness, and execution rules: omh-routing/references/skill-common-rail.md. Load it when applicable; otherwise name an unavailable capability.
1---2name: omh-memory-sync3description: [omh] English-canonical Hermes memory-review guidance: inspect USER.md and MEMORY.md claims and prepare a native write diff without invoking, applying, or observing a native write; for a new fact use memory-new, and for a past decision use decision-recall. Use when the user says: memory-sync, memory curation, memory review, memory inspect, memory check, memory update, context cleanup, curate memory.4---5
6# Memory Sync
7
8This is a Hermes-native `memory-sync` workflow skill.
9
10## Why This Exists
11
12`memory-sync` exists so Hermes users can ask for this workflow in chat and receive a structured, evidence-bounded OMH operating surface instead of ad hoc narration.
13
14## Do Not Use When
15
16- The request is already handled by a narrower explicit skill with stronger evidence.
17- The user asks OMH to secretly run external platforms, connectors, schedulers, file exports, or runtime agents.
18- The only safe answer is to ask for missing authority, credentials, target, or observed evidence first.
19
20## Examples
21
22Good example:
23
24- Prompt: memory-sync inspect stale MEMORY.md claims, prepare a native write diff, and ask which claims to keep, revise, or archive.
25- Expected behavior: Produce `prepare_memory_sync` with required context, wrapper actions, and not-evidence boundaries.
26- Why: The prompt names a real workflow surface that Hermes can orchestrate without hiding execution.
27
28Bad example:
29
30- Prompt: memory-sync claim a prepared native diff changed MEMORY.md or USER.md.
31- Expected behavior: Report the missing observed evidence or authority instead of claiming the external step happened.
32- Why: Prepared OMH guidance is not platform, runtime, connector, file, memory, or delivery evidence.
33
34## Completion Checklist
35
36- Confirm the workflow target, evidence boundary, and stop condition are named.
37- Report which outputs are prepared, observed, blocked, or missing.
38- Name the smallest next verification or handoff instead of claiming completion from narration.
39
40## Recovery Notes
41
42- If required context is missing, ask one blocking question or route back to the narrower workflow.
43- If runtime or wrapper evidence is unavailable, keep the status as not_observed and expose the next observable action.
44
45## Workflow Lane
46
47- Current lane: **Retained knowledge** (`memory-new`, `memory-sync`, `decision-recall`, `wiki`) - memory, rejected alternatives, wiki notes, retrieval, and staleness.
48- If intent belongs to another lane, hand back to `oh-my-hermes` or name the adjacent workflow.
49- Shared product, routing, compatibility, and evidence rules: `omh-routing/references/skill-common-rail.md`.
50
51## English-Canonical Interview Protocol
52
53- **Inventory (목록)** - Nobody hands you the material: call `omh_memory` with `action="status"` to get the entry inventory - per-file entry counts, per-entry index and size, headroom, and which entries have no OMH record. It returns counts and hashes, never entry text, because the text is already yours.
54- **Claim extraction (추출)** - Break the `USER.md` and `MEMORY.md` material into claims. Quote only observed claims; never invent provenance.
55- **Provenance (출처)** - Ask for the source class and distinguish Hermes-native, provider, and vector material as `not_omh_reviewed`.
56- **Target (대상)** - Review existing native-memory claims only. Route a new project/product fact to `memory-new`.
57- **Candidate selection (후보)** - Default to a short interview, not a census: pick about five candidate entries per pass and say why each was picked. Rank by the signals already on hand — dreaming reminders (duplicate clusters, deadline, `stale_review_required`), the status bridge's OMH-record similarity rows (near-duplicates), and claims that look stale, conflicting, or overgeneralized. Walk the complete inventory only when the user asks for a full review.
58- **Per-entry confirmation (확인)** - Walk the selected candidates one at a time. For each entry, quote it back from your own memory file and state what you take it to mean, then ask the user to keep, revise, or archive it before moving on. Do not summarize the whole file and ask one question about all of it; a review the user cannot correct entry by entry is not a review.
59- **Cursor (이어하기)** - Close every pass by naming what was covered and what was not — reviewed entry indexes, remaining candidates, and the next entry a resumed review would start from — so an interrupted interview resumes instead of restarting. The resume point lives only in the conversation; no store persists it, so name it explicitly rather than assuming the system remembers.
60- **Review (검토)** - Prioritize stale, conflicting, duplicate, and overgeneralized claims. Offer keep, revise, or archive choices; do not describe an archive as removal.
61- **Attention (주의)** - For a reviewed OMH-local record, keep/archive is an attention tier: `active` leads the working context, `reference` stays recallable behind active peers, `archive` leaves default recall. Preview with `omh memory attention <record-id> --tier <tier>`, say which records stay in the working context and which leave it, then apply with `--apply` only after the user agrees. The preview writes nothing.
62- **Diff (차이)** - Prepare one concise native write diff with before/after claims and counts. Keep the caps: MEMORY.md about 2,200 characters and USER.md about 1,375 characters.
63- **Native-write boundary (쓰기)** - OMH prepares guidance and a native write diff only; no OMH surface invokes, applies, or observes a `MEMORY.md`/`USER.md` write.
64- **Apply after approval (적용)** - The interview does not end at a diff. Per-entry keep/revise/archive answers are input to the diff, not approval of it: ask for one explicit approval of the assembled diff. After that approval, apply the approved entries yourself through the Hermes-native memory tool — the same tool that owns these files — and report what the write observably changed; with the tool available, leaving an approved diff silently unapplied fails the interview. If the native memory tool is unavailable, report the approved diff and stop: that is a completed review with the write pending, never a failed interview and never a reason to edit the files directly. The OMH artifact stays `memory_curation_review/v1` metadata either way: the native write is Hermes's own act and never becomes OMH mutation evidence.
65
66## Memory Boundaries
67
68The prepared artifact is `memory_curation_review/v1`, not native-memory mutation evidence. Hermes-native and external provider/vector context is `not_omh_reviewed`: it can nominate an OMH candidate but never inherits OMH approval. A configured Hermes runtime may transmit rendered OMH prefetch content in its model request.
69
70Use lifecycle words literally: expire removes influence only; retire archives recoverably; restore creates a new pending revision while preserving the archive; prune hard-deletes only the manifest-declared OMH-local target set. Report restore and prune first. No lifecycle result proves anything outside that named local target set.
71
72An attention tier is not a lifecycle state, and the two uses of "archive" are different: the `archive` tier only stops a record from occupying the default working context, leaving it in the store, readable, and answerable by `omh memory recall --include-archived`, while `retire` moves an expired revision into the local archive directory. Neither is deletion; never describe either as one.
73
74Legacy v1 material is migration/review-required. Present `memory inventory` counts and the report-first per-artifact `memory reactivate ... --apply` path; inventory and reactivation never silently grant replay eligibility.
75
76Dreaming runs automatically in reminder mode at five scheduler points: `turn` when the interval is due (default five turns), `compaction` before compression discards messages, `session_end` after a productive session, `shutdown` as the final process opportunity, and `session_start_recovery` when the prior session ended without consolidation. It prepares reminders for duplicate clusters, records at or near their deadline, headroom below the configured floor, `stale_review_required`, and `expired_volatile_records`; an unchanged standing condition is suppressed until its value changes. Anything whose source OMH cannot explain is not a candidate. Dreaming never invokes a model or performs consolidation, retirement, restore, or prune.
77
78Treat ranking signals within their limits: pins guarantee inclusion but never override expiry, scope, perspective, or review eligibility; attention tiers control working-context occupancy, not truth; `approved_manual` has 100% veracity weight and `approved_auto_safe` 90%, while an unknown approval mode fails closed to the lower weight; age only breaks ties within an equal relevance rank; and usage uses saturating buckets so repeated delivery cannot compound into a permanent lead.
79
80Normal users use natural-language Hermes chat. `omh memory ...` commands are agent/operator control-plane references, not normal-user setup.
81
82## Use When
83
84Use when existing Hermes USER.md, MEMORY.md, or accumulated skill memories need an English-canonical, claim-by-claim review. It prepares native write guidance only; no OMH surface invokes, applies, or observes a native write — a user-approved diff is applied by Hermes's own native memory tool. Do not use for new project or product candidates.
85
86 Strong routing signals: `memory-sync`, `memory curation`, `memory review`, `memory inspect`, `memory check`, `memory update`, `context cleanup`, `curate memory`, `stale memory`, `hermes remembers`, `conflicting memory`, `duplicate skill`, `MEMORY.md`, `USER.md`, `what you remember about me`, `your memory about me`, `your memories`, `memory interview`, `memories still true`, `기억하고 있는`, `기억하고 있는 프로젝트 맥락`, `기억하는 맥락`, `현재 hermes가 기억하는 맥락`, `현재 헤르메스가 기억하는 맥락`, `헤르메스가 기억하는 맥락`, `오래된 맥락`, `오래된 기억`, `기억 점검`, `기억 정리`, `메모리 업데이트`, `메모리 검사`, `메모리 점검`, `메모리 정리`, `맥락 점검`, `맥락 정리`, `맥락 피드백`, `등록된 맥락`, `헤르메스 기억`, `중복 스킬`, `나에 대해 잘못 알고`, `저장된 내 정보`, `너한테 저장된`, `저장된 프로필`, `기억 바로잡`, `메모리 인터뷰`, `기억 인터뷰`
87
88## Catalog Metadata
89
90Category: `memory`
91Phase: `curation-review`
92Hermes role: `memory-keeper`
93Quality tier: `workflow-surface-gated`
94Reasoning demand: `light`
95
96Quality bar:
97
98- Name the user-facing workflow objective, required context, next action, and stop condition.
99- Separate prepared guidance from observed platform, runtime, connector, file, memory, or delivery evidence.
100- Expose missing tools, credentials, targets, or observations as user-visible gaps.
101- State that Hermes-native and external provider/vector context is not_omh_reviewed, can nominate a candidate only, and may receive rendered OMH prefetch content through a configured Hermes runtime model request.
102
103Handoff policy:
104
105Keep this as Hermes-facing orchestration guidance first. Prepare executor, connector, gateway, or host-runtime handoff only when the user accepts that next step and observed evidence can be recorded.
106
107Required inputs:
108
109- user request
110- target context
111- delivery or status expectation
112- known missing evidence
113
114Expected outputs:
115
116- memory-sync/v1 card or guidance
117- next action
118- prepared-vs-observed boundary
119
120Artifact expectations:
121
122- memory-sync/v1 metadata-only runtime or wrapper card when recorded
123
124Safety rules:
125
126- A memory-sync review is prompt guidance only: no OMH surface invokes, applies, or observes a MEMORY.md or USER.md write. Applying a user-approved diff is Hermes's own act through its native memory tool — report it as an observed native write, never as OMH mutation evidence. Hermes-native and external provider/vector context is not_omh_reviewed and never inherits OMH approval.
127- Do not claim connector, gateway, runtime, file generation, memory mutation, or host automation evidence from prepared guidance.
128- Keep English as the canonical protocol; Korean routing triggers and concise Korean help labels remain available.
129- Quote claims only when observed, do not invent provenance, and keep the prepared native diff separate from any native write.
130
131## Runtime Evidence
132
133Preferred harness for this skill: `memory-sync`.
134
135```sh
136omh runtime record --skill memory-sync --harness memory-sync --status started
137```
138
139Record observed delegation results; otherwise return `not_available` or `not_observed`.
140Prepared OMH routing is not execution, review, CI, merge-readiness, or merge evidence.
141- Treat wrapper memory/context summaries as advisory local context, not proof of opaque Hermes memory reads or changes.
142Preserve workflow intent and stop conditions; verify before claiming completion.
143
144Use Hermes-native subagent/delegation features when available: native subagents -> Hermes delegation when available, otherwise sequential lanes.
145
146Shared product, compatibility, topology, memory, harness, and execution rules: `omh-routing/references/skill-common-rail.md`. Load it when applicable; otherwise name an unavailable capability.
rlaope