hmbown/faerie-fire

Faerie Fire

Make hidden things visible by highlighting, tagging, and marking them for attention.

What This Skill Does

Faerie Fire marks things that are already known. It does not search, discover, explain, translate, or fix. The user has already identified what matters; your job is to make those items impossible to overlook by annotating, tagging, highlighting, or visually surfacing them in context. In this grimoire, Faerie Fire is treated as a metaphorical spell with a shipping-now delivery profile. Canonical reference input: Faerie Fire (spell).

When To Use

• "Highlight", "tag", "mark", "annotate", "call out", "flag" applied to specific items the user has already named or located
• "Make X visible" where X is a known set of findings, clauses, functions, tickets, or features
• "Show me where [specific thing] appears" — location surfacing, not discovery
• "Annotate", "add margin notes", "color-code", "produce a summary table of marked items"
• The user provides the target list and asks you to apply markers to a document, codebase, dataset, or diagram

Prerequisites

• No extra runtime dependencies beyond Hermes Agent and the normal toolset for this session.

Procedure

1. Restate the target, the success condition, and any no-touch boundaries before taking action.
2. Confirm the target set: Restate what the user wants marked. If the list is implicit, extract it explicitly before annotating.
3. Choose the marking strategy based on the medium:
4. Code: inline comments, // HIGHLIGHT: markers, or extracted function lists with file:line references
5. Documents: bold/underline, margin notes, or a findings table with clause references
6. Data/tables: tagged rows, conditional formatting descriptions, or filtered views
7. Diagrams: callout labels, numbered annotations, or a legend mapping marks to gaps
8. Apply markers with context: Every mark must include why the item was flagged, not just that it was flagged. A marker without rationale is noise.
9. Deliver two artifacts: the annotated source and a concise summary listing what was marked and the selection criteria.
10. Package the result as the deliverables below, with confidence, assumptions, and unresolved risk called out explicitly.

Deliverables

• An annotated version of the source material with key findings highlighted and marked.
• A summary of what was highlighted and why each item warrants attention.

Pitfalls / Guardrails

• Keep the metaphor anchored to a real mechanism instead of drifting into lore.
• Be selective. Highlighting everything is highlighting nothing. If the user gives you 200 items, ask them to narrow or apply a ranking.
• Never claim to have discovered something you were not asked to mark. Faerie Fire illuminates; it does not hunt.
• If the request mixes marking with fixing, translating, or explaining, scope yourself to marking only and note the boundary.
• Do not use for: Scanning/searching: "Find all hidden webhooks", "Scan for MCP servers" — the user is asking you to discover, not mark. Use a detection spell instead.
• Do not use for: Explaining: "Explain what this manifest does", "What does this code do?" — comprehension, not annotation.
• Do not use for: Translating: "Translate from German to English" — language conversion, not highlighting.
• Do not use for: Rewriting: "Rewrite for executives", "Simplify this report" — transformation, not marking.
• Do not use for: Fixing: "Highlight the risky parts and fix them" — Faerie Fire only marks; it does not remediate. Decline the fix portion or route it to a separate spell.
• Do not use for: Vague requests: "Highlight the important stuff" with no criteria — ask the user what "important" means before proceeding.

Verification

• Check that the result includes every deliverable promised above.
• Check that confirmed facts, assumptions, and inferences are visibly separated.
• Check that the metaphor still maps cleanly to a real operational mechanism.

Example Invocation

/faerie-fire Use \$faerie-fire to highlight the most important [findings/patterns/elements] in this [document/codebase/dataset] so they cannot be overlooked