agentleFS
tradingstrategy-aitradingstrategy-ai/web3-ethereum-defi/.claude/skills/add-curator/SKILL.md

add-curator

Skill834 starsChanged 3 days agoMITView on GitHub

Add a new vault curator to feed metadata and curator detection. Use when the user wants to add a verified curator, protocol-managed curator, or curator alias discovered from vault data.

  • Reads credentials

The file

Show its rules and text

We split this skill into the rules it gives an agent

41 rules. About 0 are common, 24 are distinctive, and 17 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈24

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 17

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    After identifying the vaults associated with the curator, run the categorise-vault-strategy skill for every newly added vault and every …

    line 24
  2. Very specific:

    Do not stop after tagging one representative vault: every vault identified as newly associated with the curator must have an evidence-based …

    line 33
  3. Very specific:

    When adding EVM address mappings, use plain lowercase string keys in tags.py (for example "0x1234..."), not HexAddress(...) constructors.

    line 38
  4. Very specific:

    If the candidate was produced by find-new-curators, open its result file first and keep the evidence trail in mind.

    line 43
  5. Distinctive:

    If curator or strategy changes require refreshing existing cached metadata, read the migration documentation and add a new numbered …

    line 48
  6. Distinctive:

    When export-time resolution already picks up the change, document that no migration is needed.

    line 53
  7. Very specific:

    Use one clean, organisation-level public name for a curator.

    line 66
  8. Distinctive:

    Do not create separate curator records merely because an organisation has international entities, regional subsidiaries or locally …

    line 70
  9. Very specific:

    Keep the public name, short description and curator-facing long description on the main brand.

    line 73
  10. Distinctive:

    Create separate curator records only for genuinely separate, independently branded organisations, not geographic variants of one brand.

    line 81
  11. Distinctive:

    Read and follow the shared description writing style before writing or editing public descriptions.

    line 86
  12. Distinctive:

    Use clear, neutral language to explain the curator's organisation, investment remit and the protocols, strategies or stablecoins it …

    line 92
  13. Distinctive:

    Do not imply custody, discretionary control, affiliation, endorsement or a performance record unless an authoritative source supports it.

    line 96
  14. Distinctive:

    Add inline Markdown links in the body: link the official website on first mention, and link the announcement, documentation or terms that …

    line 100
  15. Distinctive:

    Run the existing curator inventory and search for nearby entries:

    line 115
  16. Distinctive:

    Check whether the same organisation already exists as a protocol or stablecoin feeder:

    line 122
  17. Very specific:

    Prefer an alias YAML with canonical-feeder-id when the organisation already has feed sources under protocols/ or stablecoins/.

    line 129
  18. Distinctive:

    When an apparent candidate is a regional or international entity, search for the main brand first and extend its protocol-specific aliases …

    line 131
  19. Distinctive:

    Use official sources when possible:

    line 137
  20. Distinctive:

    Add descriptions from primary sources:

    line 145
  21. Distinctive:

    Avoid adding aggregator pages, unofficial social accounts, or broken RSS feeds.

    line 157
  22. Very specific:

    Always add short_description and long_description for new curator YAML files.

    line 200
  23. Very specific:

    Use other-links for evidence pages such as protocol forum announcements, documentation pages, or vault launch posts that prove the …

    line 203
  24. Distinctive:

    Do not duplicate feed sources on alias files.

    line 221
  25. Distinctive:

    When the runtime supports subagents and the user has asked for or allowed delegated work, use subagents for the logo workflow:

    line 229
  26. Very specific:

    Ask one subagent to use extract-vault-protocol-logo as a template for extracting the official curator logo from the curator website or …

    line 232
  27. Very specific:

    After original logos exist, ask another subagent to use post-process-logo to create standardised PNGs in …

    line 235
  28. Distinctive:

    If subagents are not available, run the same two skills sequentially.

    line 239
  29. Distinctive:

    Prefer official brand kits, website header logos, GitHub assets, and then Twitter/X avatars in that order.

    line 240
  30. Distinctive:

    When the curator is also a vault protocol, add the logo using the protocol slug so vault metadata and curator metadata share the same asset …

    line 248
  31. Very specific:

    Add entries to CURATOR_NAME_PATTERNS only for verified vault-name variants, short names, acronyms, or product names.

    line 259
  32. Distinctive:

    Avoid generic single-word patterns unless the word is distinctive and has been checked against the vault dataset for false positives.

    line 261
  33. Very specific:

    Add blanket protocol-managed protocols to PROTOCOL_CURATED_SLUGS.

    line 271
  34. Very specific:

    Add all protocol curator slugs to ALL_PROTOCOL_CURATOR_SLUGS if the protocol is address-scoped rather than blanket-managed.

    line 272
  35. Very specific:

    Add the display name to PROTOCOL_CURATOR_NAMES.

    line 274
  36. Very specific:

    Add address-set detection in identify_curator() only when some, but not all, vaults under the protocol are protocol-managed.

    line 275
  37. Distinctive:

    Make sure the returned slug matches the feeder slug used elsewhere.

    line 277
  38. Distinctive:

    Run the curator inventory:

    line 281
  39. Very specific:

    Run targeted tests only. Use .local-test.env for pytest:

    line 293
  40. Distinctive:

    If Python was edited, run:

    line 299
  41. Distinctive:

    If YAML feed files were edited and no specific feed tests exist, at least run a script or small import path that loads feeder metadata.

    line 306

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Add curator

This skill adds a vault curator to the repository and wires it into feed tracking and Python curator detection.

Inputs

Gather or infer these before editing:

  • Curator name
  • Curator slug, using the existing feeder slug style
  • Evidence from vault data: vault names, protocols, chains, and why this is a curator
  • Curator type: third-party curator, alias to an existing feeder, protocol-managed curator, or name-pattern update
  • Short curator description and long curator description
  • Optional website, Twitter/X, LinkedIn, blog/RSS, supporting links, and logo source

Strategy tags

After identifying the vaults associated with the curator, run the categorise-vault-strategy skill for every newly added vault and every existing vault newly covered by the curator metadata. Curator attribution does not replace strategy classification. Follow .claude/skills/categorise-vault-strategy/SKILL.md to review each vault's strategy context, update the protocol-local address-level tags.py mapping, and verify get_strategy_tags() or the native vault export returns the maintained tags.

Do not stop after tagging one representative vault: every vault identified as newly associated with the curator must have an evidence-based mapping, or must remain unmapped so its resolver returns the explicit missing-information result (None for VaultBase adapters).

When adding EVM address mappings, use plain lowercase string keys in tags.py (for example "0x1234..."), not HexAddress(...) constructors. The strategy-tag lookup helper normalises adapter addresses before reading these mappings.

If the candidate was produced by find-new-curators, open its result file first and keep the evidence trail in mind.

Persisted metadata refresh

If curator or strategy changes require refreshing existing cached metadata, read the migration documentation and add a new numbered migration. Standalone curator/tag repair scripts are legacy format and must not be created or run for these tasks. Reuse maintained resolvers through the supplied context, with a reviewed scope and reporting. When export-time resolution already picks up the change, document that no migration is needed. Include the production launcher reminder when a PR with a migration is requested; keep authoring separate from production execution.

Curator research often surfaces a curator's vaults whose price history starts weeks after they launched, because the scanner did not read them earlier. Compare each newly attributed vault's first-seen block with its first price row. When the gap matters, propose a price backfill for those vaults only, as described in the historical price backfills section of the migration guide.

Naming policy

Use one clean, organisation-level public name for a curator. Prefer the main brand name without legal, jurisdictional, regional or affiliate suffixes, for example China Asset Management, not China Asset Management (Hong Kong).

  • Do not create separate curator records merely because an organisation has international entities, regional subsidiaries or locally incorporated legal vehicles.
  • Keep the public name, short description and curator-facing long description on the main brand; use the same canonical curator slug for its international entities.
  • Preserve an upstream's exact legal or regional spelling in the applicable protocol-specific mapping field (such as asseto-role) or a detection pattern, so attribution remains accurate without exposing a cluttered name.
  • Mention a specific legal entity in the long description only when it adds material factual context; do not use it as the public curator identity.
  • Create separate curator records only for genuinely separate, independently branded organisations, not geographic variants of one brand.

Website descriptions

Read and follow the shared description writing style before writing or editing public descriptions.

The YAML short_description and long_description fields are public website copy for a general audience, not repository notes.

  • Use clear, neutral language to explain the curator's organisation, investment remit and the protocols, strategies or stablecoins it oversees when those are publicly verified. State whether the curator is independent or protocol operated when that affects how users assess responsibility.
  • Do not imply custody, discretionary control, affiliation, endorsement or a performance record unless an authoritative source supports it. If no separate curator is disclosed, say so plainly rather than inferring one from contract data.
  • Add inline Markdown links in the body: link the official website on first mention, and link the announcement, documentation or terms that support a material relationship or claim. A URL in a separate YAML field is not a replacement for the in-text source.
  • Mention a public founder, executive or well-known backer only when a primary source confirms the connection and it materially helps readers understand the organisation. Link the supporting announcement or biography; never infer an investor from an aggregator database or use the name as an endorsement.
  • Exclude project-specific logic and technical internals, including classifier decisions, scanner or adapter behaviour, ABI names, contract accessors and implementation details. Put those in code comments or technical documentation instead.

Step 1: Check existing coverage

Run the existing curator inventory and search for nearby entries:

poetry run python .claude/skills/find-new-curators/scripts/print-existing-curators.py
rg -n "{curator name}|{curator slug}" eth_defi/data/feeds eth_defi/vault/curator.py

Check whether the same organisation already exists as a protocol or stablecoin feeder:

rg --files eth_defi/data/feeds -g "*.yaml"

Prefer an alias YAML with canonical-feeder-id when the organisation already has feed sources under protocols/ or stablecoins/. When an apparent candidate is a regional or international entity, search for the main brand first and extend its protocol-specific aliases rather than adding another curator record.

Step 2: Verify identity and sources

Use official sources when possible:

  • Official website homepage
  • Twitter/X account linked from the website, written without @
  • LinkedIn company slug from the official company page URL
  • RSS or Atom feed for an official blog, only if it works
  • Documentation, forum post, or vault UI proving the organisation acts as curator

Add descriptions from primary sources:

  • short_description: one concise, audience-facing sentence. Prefer the Twitter/X or LinkedIn bio when available; otherwise use the official homepage tagline. Describe the organisation, not a single vault or token.
  • long_description: two to four Markdown-safe sentences or short paragraphs following the website-description standard above. Avoid hype, rankings, unverifiable performance claims and investment advice.
  • For alias curators, describe the curator organisation represented by the alias. Do not blindly inherit a product-only description from the canonical feeder when it would misrepresent the alias.

Avoid adding aggregator pages, unofficial social accounts, or broken RSS feeds. If a social account is inferred from search results, note that it was inferred in the final answer.

Step 3: Update feed files

For a new third-party curator, create:

feeder-id: {curator-slug}
name: {Curator name}
role: curator
website: https://example.org
short_description: {One-sentence curator description.}
long_description: |
  {Two to four sentence curator description based on official homepage,
  docs, Twitter/X bio, or LinkedIn bio.}
twitter: example
linkedin: example-company
rss: https://example.org/feed.xml

# Evidence and background references.
#
# Use this section to preserve why this organisation is treated as a
# vault curator.  Prefer primary sources: protocol forum announcements,
# protocol docs, curator launch posts, vault UI pages, or official blog
# posts.  These links are metadata only; the feed collector will not
# fetch them as post sources.
#
# Be verbose with titles so a future reviewer can understand the
# evidence without opening every link.
other-links:
  - title: Protocol forum - {Curator name} announced as curator for {vault or strategy name}
    url: https://example.org/curator-announcement

Save this as:

eth_defi/data/feeds/curators/{curator-slug}.yaml

Omit unknown optional fields instead of leaving empty keys. Always add short_description and long_description for new curator YAML files. If primary-source evidence is too weak to write them, stop and ask instead of inventing descriptions. Use other-links for evidence pages such as protocol forum announcements, documentation pages, or vault launch posts that prove the organisation acts as curator.

For an alias to an existing feeder, create identity and description metadata:

# {Curator name} curator - feeds provided by {role}/{canonical-slug}.yaml
feeder-id: {curator-slug}
name: {Curator name}
role: curator
canonical-feeder-id: {canonical-slug}
short_description: {One-sentence curator description.}
long_description: |
  {Two to four sentence curator description based on official sources.}

Do not duplicate feed sources on alias files.

Step 4: Add or reuse logos

Curator logos should follow the same repository logo conventions as vault protocol logos unless a curator-specific logo location has been introduced later.

When the runtime supports subagents and the user has asked for or allowed delegated work, use subagents for the logo workflow:

  • Ask one subagent to use extract-vault-protocol-logo as a template for extracting the official curator logo from the curator website or brand kit into eth_defi/data/vaults/original_logos/{curator-slug}/.
  • After original logos exist, ask another subagent to use post-process-logo to create standardised PNGs in eth_defi/data/vaults/formatted_logos/{curator-slug}/.

If subagents are not available, run the same two skills sequentially. Prefer official brand kits, website header logos, GitHub assets, and then Twitter/X avatars in that order. Keep the original logo source documented in the logo folder when the extraction skill creates a report.

For alias curators, reuse the canonical feeder logo unless the product has a distinct curator brand.

When the curator is also a vault protocol, add the logo using the protocol slug so vault metadata and curator metadata share the same asset path.

Step 5: Update curator detection

Edit eth_defi/vault/curator.py.

For third-party curators:

  • Usually the YAML name field is enough.
  • Add entries to CURATOR_NAME_PATTERNS only for verified vault-name variants, short names, acronyms, or product names.
  • Avoid generic single-word patterns unless the word is distinctive and has been checked against the vault dataset for false positives.
  • Sort longer or more specific patterns before short aliases when adding multiple variants for the same curator.

For a name-pattern update to an existing curator, change only CURATOR_NAME_PATTERNS and explain why no new feed file was needed.

For protocol-managed curators:

  • Add blanket protocol-managed protocols to PROTOCOL_CURATED_SLUGS.
  • Add all protocol curator slugs to ALL_PROTOCOL_CURATOR_SLUGS if the protocol is address-scoped rather than blanket-managed.
  • Add the display name to PROTOCOL_CURATOR_NAMES.
  • Add address-set detection in identify_curator() only when some, but not all, vaults under the protocol are protocol-managed.
  • Make sure the returned slug matches the feeder slug used elsewhere.

Step 6: Verify

Run the curator inventory:

poetry run python .claude/skills/find-new-curators/scripts/print-existing-curators.py

Find the relevant targeted tests:

rg -n "identify_curator|CURATOR_NAME_PATTERNS|PROTOCOL_CURATED|canonical_feeder" tests eth_defi

Run targeted tests only. Use .local-test.env for pytest:

source .local-test.env && poetry run pytest {test file or test name} -v

If Python was edited, run:

poetry run ruff format eth_defi/vault/curator.py
poetry run ruff check eth_defi/vault/curator.py

If YAML feed files were edited and no specific feed tests exist, at least run a script or small import path that loads feeder metadata. For curator YAML edits, also verify descriptions are present and load through the shared schema:

poetry run python - <<'PY'
from pathlib import Path

from eth_defi.feed.sources import load_feeder_metadata

for path in sorted(Path("eth_defi/data/feeds/curators").glob("*.yaml")):
    data = load_feeder_metadata(path)
    assert data.get("short_description"), path
    assert data.get("long_description"), path

print("ok")
PY

Step 7: Report

Finish with clean-up-for-pr at the outermost task's review handoff.

Summarise:

  • Feed files created or aliased
  • Short and long descriptions added, with source pages used
  • Detection logic changed
  • Logo files added or reused
  • Verification commands run
  • Any inferred sources or unresolved identity questions

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.