These are standard SKILL.md folders and ZIP bundles that can be loaded into an AI agent environment that supports Skills. The enablement process is as follows:
-
Ask the host environment to load the skills bundle ZIP (or folder) from its URL (
file:for local reference) or local path. -
To install the plugin, place the plugin directory in the
.agents/pluginsdirectory of your AI agent environment. Activate it by adding its metadata to themarketplace.jsonfile in the same directory. -
Test the skill using the usage example documented in the
SKILL.mdfile. -
Add it to a workflow schedule if the skill needs to run in automated form on a daily, weekly, or monthly basis.
Ask your AI agent environment to load each skill from its ZIP or directory path:
file:///path/to/ai-agent-skills/<skill-name>.zip
file:///path/to/ai-agent-skills/<skill-name>/
The agent-rdf-memory/ directory is the canonical behavioral contract.
Execute this retrieval sequence before any task:
- List
agent-rdf-memory/and subfolders (sessions/,projects/,entities/) - Read
core.ttl(user identity, output paths) - Read
preferences.ttl(operational rules) - Read
preferences.private.ttlif present (private overlay) - Read
ontology.ttl(prompt-intent routing) - Read
index.ttl(session pointers) - Follow
rdfs:seeAlsoreferences to relevanthowto/*.ttl - Reason over loaded Turtle content
- Write session memory on trigger events (see
ontology.ttl#MemoryWriteTrigger)
At session start, resolve:
- Agent environment:
{AGENT_ENV_NAME}(e.g.,claude_code,opencode) - LLM model:
{LLM_MODEL}from the system prompt - Output path: from
core.ttloutput path entries
These feed artifact routing and whoami responses.
For automatic memory injection, set up the SessionStart hook per
agent-rdf-memory/SESSION-START-HOOK.md. This reads memory files before
the first user turn.
Run whoami to confirm identity resolution and output path routing are correct.