The pattern
A skill’s procedure can include a conditional invocation of another skill: check a condition, and if the condition is met, trigger the other skill before continuing. This creates skill chains — sequences of skills that execute based on the state of the repository.
Example: interpret-first-message (step 2) checks whether the latest situational assessment predates the most recent commit. If it does, the skill invokes situational-assessment before proceeding. The decision is proceduralized: the condition is explicit, the branch is documented, and a less capable agent can follow it.
What kind of thing is this
This pattern is the practice-level mechanism by which policy 001 (progressive automation) is realized. Policy 001 prescribes the direction: “make actions accomplishable by less capable agents over time.” Conditional skill invocation is the how: raw interpretation gets replaced by documented chains of condition-checked skill calls.
In semiotic-endeavor terms: policy 001 is method (nucleus, j operator — stable commitment). Skill chaining is practice (flow, G operator — what actually happens). The flow-nucleus compatibility theorem predicts they converge: method says “proceduralize”; practice discovers the mechanism (conditional invocation); the mechanism gets formalized back into method (as specifications for how skills chain); which enables even more automation. The fixed point is a system where most work is proceduralized and the remaining interpretation handles genuinely novel situations.
This is not a goal (goals are closure conditions to achieve). It is not a new policy (policy 001 already covers the commitment). It is the emergent practice consequence of policy 001 applied to the skill system — method-practice convergence in action.
Decision protocol structure
A conditional invocation has three parts:
- Condition: a checkable state (e.g., “is the assessment older than the latest commit?”)
- Branch: what skill to invoke if the condition is met
- Continuation: how to use the invoked skill’s output in the invoking skill’s procedure
Over time, more decision points get encoded this way. Each one replaces a piece of “the agent figures it out” with “the procedure says what to check.” The aggregate effect is that the system becomes progressively more operable by less capable agents — which is exactly what policy 001 asks for.
Implications for plan 0039
Plan 0039 (work procedure with decision protocols) is about documenting the general work flow with decision forks. Each fork is a candidate for conditional skill invocation. The work procedure plan should identify which forks can be proceduralized now vs. which require judgment for the time being.