← AchriomCONTENT HISTORY

Update to Achriom

Snapshot Nov 5, 2026 · 00:16 UTC · version 2.1.1

Collection source: downloaded plugin package. These snapshots do not have a confirmed matching collection source. Differences in file lists alone do not establish changes to the package.

WHAT CHANGED

Instructions updated for add

Skill instructions — Now says: `add(what: "library", media_type, title, external_id)` with the returned external_id for an exact match. remember(media_type, title, happened, date?,...

Observed in instructions or declared skills. Runtime behavior has not been tested.

Save Achriom and check future changes in Following.

Up to 3 plugins free · no card required.

All fields and original values.

Skill instructions

Before

Sep 30, 2026 · 22:51 UTC

`add_item` with the returned external_id for an exact match. update_status(media_type, title, status) set_progress(...) # partway through a book mark_tv_watched(...) # episodes watched update_rating(...) # wh...

After

Nov 5, 2026 · 00:16 UTC

`add(what: "library", media_type, title, external_id)` with the returned external_id for an exact match. remember(media_type, title, happened, date?, rating?) # finished, started, gave up, want, again set(target: "item", media_type, it...

Compare saved observations

Download comparison JSON

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -14,7 +14,7 @@
 
 Gather everything named, including works mentioned earlier in the conversation when the user says "these" or "what I mentioned."
 
-- Three or fewer items: `lookup_item(media_type, title)` per item, then `add_item` with the returned external_id for an exact match.
+- Three or fewer items: `lookup_item(media_type, title)` per item, then `add(what: "library", media_type, title, external_id)` with the returned external_id for an exact match.
 - Four or more: `bulk_add_items`.
 
 ### Step 2: Disambiguate Only When It Matters
@@ -26,10 +26,9 @@
 If they mentioned status or feeling alongside ("finished it last week," "halfway through," "loved it"), set it in the same pass:
 
 ```
-update_status(media_type, title, status)
-set_progress(...)            # partway through a book
-mark_tv_watched(...)         # episodes watched
-update_rating(...)           # when they volunteered a verdict
+remember(media_type, title, happened, date?, rating?)   # finished, started, gave up, want, again
+set(target: "item", media_type, item_id, fields: {progress, rating})   # partway through, a verdict
+add(what: "episode", media_type: "show", item_id, season, episode, through: true)   # episodes watched
 ```
 
 ### Step 4: Confirm, Then Notice
Full technical diff · 1 changed fields

changed /skill_md_contents

BEFORE
"---\nname: \"add\"\ndescription: \"Fast library intake. Add books, movies, albums, shows, anime, or podcasts you name, with correct identification and optional status in one pass\"\n---\n\n# Build the Library Fast\n\nTurn whatever the user names into correctly identified library entries, with minimum friction and no lost momentum.\n\n\n## Workflow\n\n### Step 1: Collect and Identify\n\nGather everything named, including works mentioned earlier in the conversation when the user says \"these\" or \"what I mentioned.\"\n\n- Three or fewer items: `lookup_item(media_type, title)` per item, then `add_item` with the returned external_id for an exact match.\n- Four or more: `bulk_add_items`.\n\n### Step 2: Disambiguate Only When It Matters\n\nAsk ONE quick question when the wrong match would genuinely hurt (remakes, common titles, same-name works by different creators). Never stall a ten-item intake on one ambiguous title: add the nine, flag the one.\n\n### Step 3: Carry the Context\n\nIf they mentioned status or feeling alongside (\"finished it last week,\" \"halfway through,\" \"loved it\"), set it in the same pass:\n\n```\nupdate_status(media_type, title, status)\nset_progress(...)            # partway through a book\nmark_tv_watched(...)         # episodes watched\nupdate_rating(...)           # when they volunteered a verdict\n```\n\n### Step 4: Confirm, Then Notice\n\nConfirm compactly: what was added, one line. Then, once five or more items exist in the library, name ONE real pattern forming across them: a recurring theme, era, or sensibility. One specific observation, not a summary. This is the moment the collection starts to feel understood.\n\nNever analyze patterns from fewer than three items; there is not enough signal.\n\n## Voice\n\nEfficient and warm. The confirmation is short; the observation is the payoff.\n"
AFTER
"---\nname: \"add\"\ndescription: \"Fast library intake. Add books, movies, albums, shows, anime, or podcasts you name, with correct identification and optional status in one pass\"\n---\n\n# Build the Library Fast\n\nTurn whatever the user names into correctly identified library entries, with minimum friction and no lost momentum.\n\n\n## Workflow\n\n### Step 1: Collect and Identify\n\nGather everything named, including works mentioned earlier in the conversation when the user says \"these\" or \"what I mentioned.\"\n\n- Three or fewer items: `lookup_item(media_type, title)` per item, then `add(what: \"library\", media_type, title, external_id)` with the returned external_id for an exact match.\n- Four or more: `bulk_add_items`.\n\n### Step 2: Disambiguate Only When It Matters\n\nAsk ONE quick question when the wrong match would genuinely hurt (remakes, common titles, same-name works by different creators). Never stall a ten-item intake on one ambiguous title: add the nine, flag the one.\n\n### Step 3: Carry the Context\n\nIf they mentioned status or feeling alongside (\"finished it last week,\" \"halfway through,\" \"loved it\"), set it in the same pass:\n\n```\nremember(media_type, title, happened, date?, rating?)   # finished, started, gave up, want, again\nset(target: \"item\", media_type, item_id, fields: {progress, rating})   # partway through, a verdict\nadd(what: \"episode\", media_type: \"show\", item_id, season, episode, through: true)   # episodes watched\n```\n\n### Step 4: Confirm, Then Notice\n\nConfirm compactly: what was added, one line. Then, once five or more items exist in the library, name ONE real pattern forming across them: a recurring theme, era, or sensibility. One specific observation, not a summary. This is the moment the collection starts to feel understood.\n\nNever analyze patterns from fewer than three items; there is not enough signal.\n\n## Voice\n\nEfficient and warm. The confirmation is short; the observation is the payoff.\n"
Full snapshot data
{
  "description": "Fast library intake. Add books, movies, albums, shows, anime, or podcasts you name, with correct identification and optional status in one pass",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 294
    }
  ],
  "name": "add",
  "skill_md_contents": "---\nname: \"add\"\ndescription: \"Fast library intake. Add books, movies, albums, shows, anime, or podcasts you name, with correct identification and optional status in one pass\"\n---\n\n# Build the Library Fast\n\nTurn whatever the user names into correctly identified library entries, with minimum friction and no lost momentum.\n\n\n## Workflow\n\n### Step 1: Collect and Identify\n\nGather everything named, including works mentioned earlier in the conversation when the user says \"these\" or \"what I mentioned.\"\n\n- Three or fewer items: `lookup_item(media_type, title)` per item, then `add(what: \"library\", media_type, title, external_id)` with the returned external_id for an exact match.\n- Four or more: `bulk_add_items`.\n\n### Step 2: Disambiguate Only When It Matters\n\nAsk ONE quick question when the wrong match would genuinely hurt (remakes, common titles, same-name works by different creators). Never stall a ten-item intake on one ambiguous title: add the nine, flag the one.\n\n### Step 3: Carry the Context\n\nIf they mentioned status or feeling alongside (\"finished it last week,\" \"halfway through,\" \"loved it\"), set it in the same pass:\n\n```\nremember(media_type, title, happened, date?, rating?)   # finished, started, gave up, want, again\nset(target: \"item\", media_type, item_id, fields: {progress, rating})   # partway through, a verdict\nadd(what: \"episode\", media_type: \"show\", item_id, season, episode, through: true)   # episodes watched\n```\n\n### Step 4: Confirm, Then Notice\n\nConfirm compactly: what was added, one line. Then, once five or more items exist in the library, name ONE real pattern forming across them: a recurring theme, era, or sensibility. One specific observation, not a summary. This is the moment the collection starts to feel understood.\n\nNever analyze patterns from fewer than three items; there is not enough signal.\n\n## Voice\n\nEfficient and warm. The confirmation is short; the observation is the payoff.\n"
}

SHA-256 of public snapshot: d9a4e057b9e2d2f33ab2380c09f3ab7a9a90f9cb26a4e7977e2063a67ea1e2f7