Agent Startup And Context Ingestion

Agent Startup and Context Ingestion

How an agent becomes useful after launch: AGENTS.md overlays, role files, skills, rig specs, workflow specs, startup checklists, refocus messages, and any future "rig context" command or skill-librarian surface.

It is the current concrete startup path inside the broader context-engineering-and-retrieval primitive. Startup gets a seat into the right initial shape; context engineering is the larger question of how a seat gets the right context for the work it is doing right now.

Most coordination failures are not tool failures; they are context failures. Agents need to know their role, operating mode, coordination convention, boundaries, and current product intent. If startup context is scattered or stale, agents execute the wrong thing very efficiently.

Use this when

• Authoring a new agent's startup files (role / culture / startup-context)
• Refreshing a seat that's been running on stale guidance
• Auditing whether agents got the current operating mode (not just what's in old files)
• Designing the orchestrator → next-agent context-transmission shape
• Building a startup map for OpenRig-building rigs

Don't use this when

• The agent is being created via Agent Starter — the starter's manifest carries startup context
• The work is artifact-backed mental-model rebuild from a packet — that's claude-compaction-restore
• The intent is to ship reusable startup content as a skill — write the skill against the agentskills.io specification, not as inline startup prose

Failure modes (4)

1. A new agent starts from an old rig spec and misses the current operating mode. Specs go stale; current state must be visible at startup, not just historical config.
2. Guidance is written to a file that future agents read, but current agents are never told. File edits don't propagate to running sessions. Cultural rollout (broadcast + fleet-changes-feed) is needed alongside file edits.
3. A startup file becomes a dumping ground and loses the map-to-canonical-sources role. Startup should point AT canonical sources; it shouldn't TRY to be one.
4. The orchestrator transmits implementation instructions without preserving product intent. Instructions decay; intent travels.

Proof standard

Startup proof should:

1. Launch or refresh a seat
2. Inspect what it actually read
3. Verify it can state the current role, mode, active constraints, and next handoff convention in its own words

Cross-runtime startup paths (5; do not collapse)

When a seat's startup is artifact-backed mental-model rebuild (active-work reentry case, distinct from reusable Agent Starters / priming packs):

• See the cross-runtime restore/reentry packet standard v0
• Source-trust ranking applies: rig whoami > target rigspec > bounded latest transcript > full transcript > touched-files > restore-summary.json

Memory surfaces consumed at startup

The primary surfaces a startup ingestion path reads:

• AGENTS / role / CULTURE / startup overlays (the per-seat identity layer)
• startup replay context (what the agent did before the current boot)
• restore / reentry packets and Agent Starters (cross-runtime startup path)

Authority-rank + read-vs-write posture governs which surfaces a startup-ingestion path can write vs only read; treat overlays as write-restricted unless your role explicitly authors them.

Startup files vs skills (the distinction)

Per agent-startup-guide.md (product reference doc) and the team handbook:

| Startup files | Skills | |---|---| | Rig-specific, role-specific identity | Reusable SOPs / methodology / knowledge | | Tell agent WHO it is, WHAT it's working on, HOW this team operates | Tell agent HOW to do something (transferable across rigs) | | Examples: role.md, CULTURE.md, startup/context.md | Examples: openrig-user, test-driven-development, dogfood | | Authored per-rig | Authored once, used everywhere |

Don't put skill content in startup files. Don't put identity content in skills. The 7-layer additive startup model (agent / profile / rig / culture / pod / member / operator) handles the layering.

See also

• forming-an-openrig-mental-model skill — orientation for new agents
• claude-compaction-restore skill — restore-time startup ingestion after compaction
• seat-continuity-and-handover skill — packet shape that survives seat resets
• https://agentskills.io/specification — the cross-runtime skill standard (what skill content should and shouldn't be)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: mvschwarz
• Source: mvschwarz/openrig
• License: Apache-2.0
• Homepage: https://openrig.dev

Install and usage instructions live in the source repository linked above.