The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted meeting transcript data from external providers (Fireflies.ai) that could contain malicious instructions.
Ingestion points: Meeting transcripts are fetched via the FirefliesClient.get_transcript() method as described in SKILL.md (Step 3).
Boundary markers: The skill does not employ specific delimiters or "ignore instructions" warnings when processing the transcript text into structured fields.
Capability inventory: The skill has permissions to write files to docs/discussions/ and .claude/data/discussions/, and it can trigger the /secondbrain-adr skill workflow.
Sanitization: The instructions do not specify any sanitization, escaping, or validation of the external transcript content before it is processed or written to disk.
[DYNAMIC_EXECUTION]: The skill uses dynamic module loading from local paths based on the provider configuration.
Evidence: SKILL.md instructs the agent to modify sys.path and import a library whose name is determined by the integrations.transcription.provider setting (e.g., from fireflies import FirefliesClient).
[EXTERNAL_DOWNLOADS]: The skill fetches data from Fireflies.ai, a well-known meeting transcription service.
Details: Accesses meeting data and summaries from https://api.fireflies.ai/graphql using a Bearer token.
[COMMAND_EXECUTION]: The skill can trigger other agent capabilities based on the data it processes from the transcripts.
Evidence: In Step 7, the skill scans decisions for potential ADR candidates and triggers the /secondbrain-adr skill workflow if the user confirms the action.

secondbrain-transcribe