# Shaping Plan > Design a shaping procedure to build complex behavior through successive approximations, specifying starting behavior, terminal behavior, and reinforcement steps. - Skill: `sethmblack/shaping-plan` (Agent Skill) - Install (CLI): `npx skillmds add sethmblack/shaping-plan` - Raw SKILL.md: https://api.skillmd.com/api/skills/sethmblack/shaping-plan/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - License: MIT - Author: sethmblack (https://skillmd.com/u/sethmblack) - Updated: 2026-09-08 - Page: https://skillmd.com/skills/sethmblack/shaping-plan --- # Shaping Plan Design a shaping procedure to build complex behavior through successive approximations, specifying starting behavior, terminal behavior, and reinforcement steps. **Token Budget:** ~700 tokens --- ## Constraints - **Do not expect perfection immediately.** Shaping requires patience and small steps. - **Do not create ratio strain.** Steps must be small enough to maintain behavior. - **Do not skip reinforcement.** Each approximation must be reinforced before raising criteria. --- ## When to Use Invoke this skill when: - Teaching a new skill or behavior - Someone asks "How do I train someone to do X?" - Current behavior is far from desired behavior - Someone says "They can't do the whole thing yet" - Building habits or behavioral repertoires - Someone asks "How do I get from here to there?" --- ## Inputs | Input | Required | Description | |-------|----------|-------------| | Terminal behavior | Yes | The final desired behavior (precisely defined) | | Starting behavior | Yes | What the person/team can currently do | | Context | No | The situation where behavior should occur | | Timeline | No | How quickly behavior needs to develop | | Available reinforcers | No | What can be used to reinforce approximations | --- ## Workflow ### Step 1: Define Terminal Behavior Precisely What does success look like? Make it observable and measurable. - Not: "Good communication skills" - Instead: "Provides status updates in standup with problem/action format in under 2 minutes" ### Step 2: Identify Starting Behavior What does the person already do that moves toward the goal? - What approximation is currently in their repertoire? - What related behaviors exist? - Where is the natural starting point? ### Step 3: Map the Approximation Path Identify 3-8 successive approximations between start and terminal. - Each step should be achievable given the previous step - Steps should be small enough that behavior is maintained - Steps should be large enough that progress is visible ### Step 4: Specify Reinforcement for Each Step For each approximation: - What reinforcer will be used? - When will it be delivered (immediately after behavior) - What schedule (continuous during acquisition) - What is the criterion for moving to next step? ### Step 5: Plan for Schedule Transition After terminal behavior is acquired: - Move from continuous to intermittent reinforcement - Specify the fading schedule - Identify natural reinforcers to maintain behavior --- ## Output Format ```markdown ## Shaping Plan **Terminal Behavior:** [precisely defined final behavior] **Starting Behavior:** [what is currently in repertoire] ### Approximation Sequence | Step | Approximation | Criterion for Advancement | Reinforcer | |------|---------------|---------------------------|------------| | 1 | [first approximation] | [when to move on] | [what reinforcer] | | 2 | [second approximation] | [when to move on] | [what reinforcer] | | ... | ... | ... | ... | | N | [terminal behavior] | [maintenance criterion] | [natural reinforcer] | ### Reinforcement Specifications - **Acquisition Schedule:** Continuous (every correct response) - **Advancement Criterion:** [number of successful responses before raising criterion] - **Maintenance Schedule:** [schedule after terminal behavior is acquired] ### Ratio Strain Prevention - **Warning Signs:** [what indicates criteria are too demanding] - **Recovery Procedure:** [how to recover if ratio strain occurs] ### Timeline Estimate [Estimated time to reach terminal behavior, with caveats] ``` --- ## Outputs **Primary Output:** A structured analysis document that identifies and articulates patterns, insights, and actionable recommendations based on the input data. **Format:** ```markdown ## Analysis: [Topic] ### Key Findings - [Finding 1] - [Finding 2] - [Finding 3] ### Recommendations 1. [Action 1] 2. [Action 2] 3. [Action 3] ``` **Example output:** See the Example section below for a complete demonstration. ## Example **Input:** "I want to train a junior developer to conduct code reviews independently. Currently they just look at code and say 'looks good.'" **Output:** **Why this works:** This example demonstrates the key principles of the skill in action. The approach is effective because: - It follows the systematic workflow outlined above - It shows concrete application of the framework - It produces actionable, specific outputs rather than vague generalizations - The analysis is grounded in observable details - The recommendations are prioritized and implementable **Alternative applications:** This same approach can be applied to: - Different contexts within the same domain - Related but distinct problem types - Scaled up or down depending on scope - Combined with complementary analytical frameworks ## Shaping Plan **Terminal Behavior:** Conducts independent code reviews that identify at least one substantive issue per review and provides actionable feedback with code suggestions **Starting Behavior:** Looks at code, provides non-specific approval ("looks good") ### Approximation Sequence | Step | Approximation | Criterion for Advancement | Reinforcer | |------|---------------|---------------------------|------------| | 1 | Identifies any single observation about the code (even obvious or stylistic) | 3 consecutive reviews with at least 1 observation | Verbal acknowledgment + "Good observation" | | 2 | Identifies one potential issue or question per review | 3 consecutive reviews with substantive observation | Acknowledgment + discussion of the observation | | 3 | Provides specific line reference with observation | 3 reviews with line-specific feedback | Senior review of their review + feedback | | 4 | Suggests a specific change or asks a clarifying question | 3 reviews with actionable feedback | Public recognition in team channel | | 5 | Identifies multiple issues per review (2+) | 5 reviews meeting criterion | Reduced oversight, increased autonomy | | 6 | Conducts independent review with actionable feedback on all substantive issues | Consistent performance over 10 reviews | Natural reinforcer: influence on codebase, team trust | ### Reinforcement Specifications - **Acquisition Schedule:** Continuous - acknowledge every approximation during early steps - **Advancement Criterion:** 3 successful responses before raising criterion (prevents accidental advancement) - **Maintenance Schedule:** Variable ratio (approximately every 3rd good review receives explicit acknowledgment) ### Ratio Strain Prevention - **Warning Signs:** Junior dev avoiding code reviews, producing lower quality observations than before, expressing frustration - **Recovery Procedure:** Drop back one step, reinforce current approximation until strong, then advance more gradually ### Timeline Estimate Approximately 4-6 weeks to reach terminal behavior, assuming 3-5 code reviews per week. Individual variation expected; some may progress faster if starting behavior is stronger than assessed. --- ## Common Mistakes to Avoid 1. **Steps too large:** If behavior deteriorates, steps are too big 2. **Advancing too quickly:** Reinforce each approximation until strong 3. **Inconsistent reinforcement:** During acquisition, reinforce every correct response 4. **No clear criteria:** Specify exactly when to advance 5. **Forgetting to fade:** After acquisition, transition to intermittent schedule --- ## Integration Use after **contingency-analysis** has identified that the behavior does not exist and must be built. Shaping builds new repertoires; it does not fix existing contingencies. --- ## Source Expert Derived from B.F. Skinner's shaping through successive approximations. "You cannot reinforce behavior that does not occur—you must shape it."