learner

Name: learner
Author: tranhieutt

69stars41forks

How to install learner?

tranhieutt/software_development_department/learner

$npx -y skills add tranhieutt/software_development_department --skill learner

Installs into the current project.

Summary

Turns real agent failures, repeated prompts, team-specific workflows, and durable project lessons into better SDD skills or memory entries. Use when the user asks to create/update/refine skills, extract reusable lessons, improve skill routing, encode team process, or save pattern

Files · 1

View on GitHub

SKILL.md

1	# Learner
2
3	Convert real SDD usage into durable operating knowledge.
4
5	Use this workflow to decide whether a lesson belongs in:
6
7	- an existing skill,
8	- a new skill,
9	- `.claude/memory/annotations.md`,
10	- another Tier 2 memory file,
11	- or no durable artifact.
12
13	## Source Principles
14
15	This workflow adopts the Agent Skills guidance that durable skills come from
16	real expertise, project artifacts, execution traces, and repeated refinement.
17	Skills should capture team-specific process and failure modes, not generic
18	best practices or deterministic glue better handled by scripts, hooks, or MCP.
19
20	## Extraction Gate
21
22	Create or modify a skill only when at least one signal is present:
23
24	- Agent made a wrong choice despite a correct prompt.
25	- Existing skill fired but failed its mission.
26	- Needed skill did not fire because `description` was weak or absent.
27	- A teammate/user wrote the same long prompt, plan, or checklist a second time.
28	- Session repeated a costly investigation, setup, verification, or handoff loop.
29	- User corrected a project convention, team preference, or non-obvious edge case.
30	- Internal process, internal system, or proprietary data pattern must be reused.
31
32	Do not create a skill for:
33
34	- General advice the model already knows.
35	- One-off code snippets.
36	- Secrets, credentials, or environment-specific auth hacks.
37	- Simple deterministic checks better implemented as hooks, scripts, tests, or MCP.
38	- Large copied docs without a clear load condition.
39
40	## Decision
41
42	Classify the lesson before editing:
43
44	\| Lesson type \| Target \|
45	\| --- \| --- \|
46	\| Existing workflow missed a rule, edge case, or output shape \| Update that skill body \|
47	\| Existing skill should have fired but did not \| Tighten that skill `description` \|
48	\| Repeated team-specific process forms a coherent unit \| Create or update a skill \|
49	\| Non-obvious caveat tied to a service/library/repo area \| Use `annotate` memory \|
50	\| Broad preference or project operating rule \| Update the right Tier 2 memory/doc \|
51	\| Deterministic repeated operation \| Prefer tested script/hook/MCP; skill only orchestrates when needed \|
52
53	Prefer updating an existing skill over adding a new one when the lesson fits an
54	existing coherent workflow.
55
56	## Workflow
57
58	1. Gather evidence:
59	- original prompt or task,
60	- correction or failure mode,
61	- first point where the agent went wrong,
62	- skills that fired or failed to fire,
63	- files, commands, traces, review comments, or user preferences involved.
64	2. Choose the smallest durable target using the Decision table.
65	3. Edit with progressive disclosure:
66	- keep `SKILL.md` under 500 lines when practical,
67	- keep only always-needed instructions in `SKILL.md`,
68	- move long examples, schemas, or domain references into `references/`,
69	- state exactly when to read each reference file.
70	4. Tune invocation:
71	- put trigger phrases and scope in `description`,
72	- add exclusions when false positives are likely,
73	- avoid relying on body-only "when to use" text for activation.
74	5. Preserve SDD gates:
75	- do not weaken `using-sdd`, permissions, hooks, or source-of-truth rules,
76	- route spec, plan, code, review, and release changes through their owning skills.
77	6. Validate:
78	- run `powershell -ExecutionPolicy Bypass -File scripts\validate-skills.ps1`,
79	- run `node scripts\validate-readme-sync.js` if counts or README inventory changed,
80	- run `node scripts\harness-audit.js --compact` for routing, hook, or harness changes.
81
82	## Skill Edit Rules
83
84	- Add what the agent lacks; cut what generic model knowledge already covers.
85	- Prefer procedures over declarations.
86	- Use defaults, not broad menus of equal options.
87	- Make fragile operations prescriptive; leave flexible judgment where multiple approaches are valid.
88	- Add concrete gotchas where the agent is likely to make the wrong assumption.
89	- Keep bundled scripts deterministic and tested.
90	- Treat each skill like a function: one coherent responsibility, composable with other skills.
91
92	## New Skill Template
93
94	Use this shape when a new SDD skill is justified:
95
96	````markdown
97	---
98	name: short-action-name
99	type: workflow
100	description: "What this skill does. Use when <specific trigger phrases, task contexts, and boundaries>."
101	argument-hint: "[expected input]"
102	user-invocable

People who liked this also viewed

6 picks

Type

learner

69stars41forks

View on GitHub

How to install learner?

tranhieutt/software_development_department/learner

$npx -y skills add tranhieutt/software_development_department --skill learner

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Learner
2
3	Convert real SDD usage into durable operating knowledge.
4
5	Use this workflow to decide whether a lesson belongs in:
6
7	- an existing skill,
8	- a new skill,
9	- `.claude/memory/annotations.md`,
10	- another Tier 2 memory file,
11	- or no durable artifact.
12
13	## Source Principles
14
15	This workflow adopts the Agent Skills guidance that durable skills come from
16	real expertise, project artifacts, execution traces, and repeated refinement.
17	Skills should capture team-specific process and failure modes, not generic
18	best practices or deterministic glue better handled by scripts, hooks, or MCP.
19
20	## Extraction Gate
21
22	Create or modify a skill only when at least one signal is present:
23
24	- Agent made a wrong choice despite a correct prompt.
25	- Existing skill fired but failed its mission.
26	- Needed skill did not fire because `description` was weak or absent.
27	- A teammate/user wrote the same long prompt, plan, or checklist a second time.
28	- Session repeated a costly investigation, setup, verification, or handoff loop.
29	- User corrected a project convention, team preference, or non-obvious edge case.
30	- Internal process, internal system, or proprietary data pattern must be reused.
31
32	Do not create a skill for:
33
34	- General advice the model already knows.
35	- One-off code snippets.
36	- Secrets, credentials, or environment-specific auth hacks.
37	- Simple deterministic checks better implemented as hooks, scripts, tests, or MCP.
38	- Large copied docs without a clear load condition.
39
40	## Decision
41
42	Classify the lesson before editing:
43
44	\| Lesson type \| Target \|
45	\| --- \| --- \|
46	\| Existing workflow missed a rule, edge case, or output shape \| Update that skill body \|
47	\| Existing skill should have fired but did not \| Tighten that skill `description` \|
48	\| Repeated team-specific process forms a coherent unit \| Create or update a skill \|
49	\| Non-obvious caveat tied to a service/library/repo area \| Use `annotate` memory \|
50	\| Broad preference or project operating rule \| Update the right Tier 2 memory/doc \|
51	\| Deterministic repeated operation \| Prefer tested script/hook/MCP; skill only orchestrates when needed \|
52
53	Prefer updating an existing skill over adding a new one when the lesson fits an
54	existing coherent workflow.
55
56	## Workflow
57
58	1. Gather evidence:
59	- original prompt or task,
60	- correction or failure mode,
61	- first point where the agent went wrong,
62	- skills that fired or failed to fire,
63	- files, commands, traces, review comments, or user preferences involved.
64	2. Choose the smallest durable target using the Decision table.
65	3. Edit with progressive disclosure:
66	- keep `SKILL.md` under 500 lines when practical,
67	- keep only always-needed instructions in `SKILL.md`,
68	- move long examples, schemas, or domain references into `references/`,
69	- state exactly when to read each reference file.
70	4. Tune invocation:
71	- put trigger phrases and scope in `description`,
72	- add exclusions when false positives are likely,
73	- avoid relying on body-only "when to use" text for activation.
74	5. Preserve SDD gates:
75	- do not weaken `using-sdd`, permissions, hooks, or source-of-truth rules,
76	- route spec, plan, code, review, and release changes through their owning skills.
77	6. Validate:
78	- run `powershell -ExecutionPolicy Bypass -File scripts\validate-skills.ps1`,
79	- run `node scripts\validate-readme-sync.js` if counts or README inventory changed,
80	- run `node scripts\harness-audit.js --compact` for routing, hook, or harness changes.
81
82	## Skill Edit Rules
83
84	- Add what the agent lacks; cut what generic model knowledge already covers.
85	- Prefer procedures over declarations.
86	- Use defaults, not broad menus of equal options.
87	- Make fragile operations prescriptive; leave flexible judgment where multiple approaches are valid.
88	- Add concrete gotchas where the agent is likely to make the wrong assumption.
89	- Keep bundled scripts deterministic and tested.
90	- Treat each skill like a function: one coherent responsibility, composable with other skills.
91
92	## New Skill Template
93
94	Use this shape when a new SDD skill is justified:
95
96	````markdown
97	---
98	name: short-action-name
99	type: workflow
100	description: "What this skill does. Use when <specific trigger phrases, task contexts, and boundaries>."
101	argument-hint: "[expected input]"
102	user-invocable

People who liked this also viewed

6 picks

Type