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

document-curator-incident

Skill834 starsChanged 3 days agoMITView on GitHub

Document curator incidents from one or more post links in curator YAML metadata, including affected vault addresses and protocols. Use when the user supplies social posts, announcements, post-mortems, or reports about a vault curator and wants to add a new incident or merge evidence into an existing incident.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 8 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When a social site blocks direct access, use indexed search results or another public copy of the same post.

    line 27
  2. Very specific:

    Use the publication date of the primary, most authoritative post as the incident date in YYYY-MM-DD format.

    line 32
  3. Distinctive:

    Search existing curator metadata before editing:

    line 39
  4. Very specific:

    If no curator YAML exists, use the add-curator skill first.

    line 50
  5. Very specific:

    Read the curator's existing incidents list in full.

    line 55
  6. Distinctive:

    Treat an overlapping link as definitive evidence that the record already exists.

    line 59
  7. Distinctive:

    Add a new record when the posts concern a separate event.

    line 60
  8. Distinctive:

    If the supplied links cover multiple unrelated events, create or merge one record per event rather than combining them.

    line 61
  9. Distinctive:

    When merging, append only new links, preserve useful existing evidence, and revise the title, description, date, label, or severity only …

    line 64
  10. Distinctive:

    Use this YAML shape:

    line 71
  11. Distinctive:

    Write a neutral title and a three-to-five-sentence Markdown description.

    line 91
  12. Very specific:

    Use the incident's main nature as the label.

    line 108
  13. Very specific:

    Do not change risk.status merely because an incident is added.

    line 114

≈ “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

Document curator incident

Turn post links into a sourced curator incident under eth_defi/data/feeds/curators/{curator-slug}.yaml.

Input

Require one or more post links. Ask for replacement links only when none of the supplied pages or indexed copies provide enough evidence to identify the curator and describe the incident without guessing.

Step 1: Read and verify the posts

Open every supplied link and record:

  • The author and publication date
  • The curator named or clearly implicated
  • The underlying event, impact, and current resolution status
  • The specifically affected vault addresses and protocol slugs
  • Whether statements are first-party facts, a post-mortem, or third-party claims

When a social site blocks direct access, use indexed search results or another public copy of the same post. Do not treat an inaccessible page as evidence and do not turn an unverified allegation into a statement of fact. Attribute disputed or third-party claims in the incident description.

Use the publication date of the primary, most authoritative post as the incident date in YYYY-MM-DD format. If an authoritative source states a distinct event date, prefer the event date. When merging, retain the existing date unless the new evidence establishes a more accurate one.

Step 2: Identify the curator YAML

Search existing curator metadata before editing:

rg -n -i "{curator name}|{curator handle}" eth_defi/data/feeds/curators eth_defi/vault/curator.py

Match the organisation that managed or curated the affected vault. Do not assign the incident to a protocol, asset issuer, quoted commentator, or reposting account merely because it appears in a post. If more than one existing curator is plausible, stop and ask the user which curator is intended.

If no curator YAML exists, use the add-curator skill first. Do not create an incomplete curator record as part of this workflow.

Step 3: Decide whether to add or merge

Read the curator's existing incidents list in full.

  • Merge when a record describes the same underlying event, even when the title, post date, or source URL differs.
  • Treat an overlapping link as definitive evidence that the record already exists.
  • Add a new record when the posts concern a separate event.
  • If the supplied links cover multiple unrelated events, create or merge one record per event rather than combining them.

When merging, append only new links, preserve useful existing evidence, and revise the title, description, date, label, or severity only when the new sources justify the change. Merge newly confirmed vault addresses and protocols into their existing lists without removing previously sourced context.

Step 4: Classify and write the incident

Use this YAML shape:

incidents:
  - date: YYYY-MM-DD
    links:
      - https://example.com/primary-source
      - https://example.com/additional-source
    vault_addresses:
      - 0x1234567890abcdef1234567890abcdef12345678
    protocols:
      - lagoon-finance
    title: Concise incident title
    description: |
      Write three to five factual Markdown sentences with [inline source links](https://example.com/primary-source).
      Attribute claims and use [additional sources](https://example.com/additional-source) where they support the body text.
    incident_kind: misleading
    severity: other

Write a neutral title and a three-to-five-sentence Markdown description. Every description paragraph must contain at least one inline Markdown link. Prefer links embedded on the claim they support; do not use bare URLs or a separate sources list inside the description. Avoid promotional language, investment advice, and facts not supported by the supplied posts.

Choose exactly one incident_kind:

  • collapse: the curator or its managed operation failed broadly, became insolvent, or ceased functioning
  • significant_loss: investors suffered a material loss without a full collapse
  • minor_loss: losses were limited in scope or impact
  • misleading: accounting, valuation, performance, or public disclosures were materially misleading
  • questionable_behaviour: substantiated governance, conflict, or conduct concerns that fit none of the labels above

Use the incident's main nature as the label. Do not invent numerical loss thresholds. Also set the schema-required severity to collapse, significant_loss, minor_loss, or other based on demonstrated impact. A misleading or questionable_behaviour incident normally has severity other unless sources demonstrate a loss or collapse.

Do not change risk.status merely because an incident is added. Risk review is a separate decision unless the user explicitly requests it.

Step 5: Update context lists and validate

Store all supplied links that support the incident in links, preserving the primary source first and removing exact duplicates. Links must use HTTP(S).

Fill vault_addresses with only the vaults demonstrated to be affected by the sources. Resolve named vaults through repository metadata or the public vault page, and lowercase EVM addresses. Include downstream vaults when a source explicitly documents contagion through a position in the directly affected vault, but do not add every vault managed by the curator. Fill protocols with the canonical protocol slugs used by repository vault metadata. Both lists must be non-empty. If an affected address or protocol cannot be verified, stop and ask the user for the missing context rather than guessing.

Validate the edited curator through the shared strict YAML schema:

CURATOR_YAML=eth_defi/data/feeds/curators/{curator-slug}.yaml poetry run python - <<'PY'
import os
from pathlib import Path

from eth_defi.feed.sources import load_feeder_metadata

path = Path(os.environ["CURATOR_YAML"])
metadata = load_feeder_metadata(path)
assert metadata.get("incidents"), path
print(path)
PY

Then run:

git diff --check

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

Report the curator, whether the incident was added or merged, the selected date, incident_kind, severity, and the number of links, vault addresses, and protocols now attached to the record.

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.