suede-recommend-next-action

Name: suede-recommend-next-action
Author: JasonColapietro

44stars8forks

How to install suede-recommend-next-action?

jasoncolapietro/suede-creator-skills/suede-recommend-next-action

$npx -y skills add JasonColapietro/suede-creator-skills --skill suede-recommend-next-action

Installs into the current project.

Summary

Next-action selector for this pack: reads current repo, terminal, plan, or handoff state, scores 2-4 candidate moves against goal alignment, unblocking, evidence, urgency, and leverage, and returns one recommended action plus a short, self-contained copy/paste prompt — expanding

Files · 1

View on GitHub

SKILL.md

1	# Suede Recommend Next Action
2
3	Recommend one action and package it as a short runnable prompt. Inspect
4	current state read-only; do not execute the recommended action unless the
5	user separately authorizes execution. Keep the full operator contract hidden
6	until the user asks to expand it.
7
8	## Recommendation Workflow
9
10	1. Resolve the target and the user's actual done outcome from the current
11	request, conversation, handoff, plan, repo, or live surface.
12	2. Check only the evidence needed to distinguish the next move. Prefer, in
13	order: current terminal/repo/live state, current source documents, current
14	plans or handoffs, then older memory.
15	3. Generate 2-4 candidate actions internally. Exclude work already verified as
16	complete, adjacent cleanup, and actions outside the user's authorized scope.
17	4. Score each candidate from 0-2 on every criterion below. Recommend the
18	highest total.
19
20	\| Criterion \| 2 points \| 1 point \| 0 points \|
21	\|---\|---\|---\|---\|
22	\| Goal alignment \| Directly produces the user's done signal \| Required prerequisite \| Merely adjacent \|
23	\| Unblocking \| Unlocks a core path or at least two downstream steps \| Unlocks one step \| Unlocks nothing known \|
24	\| Evidence \| Confirmed by current source \| Confirmable with one read-only check \| Depends on an assumption \|
25	\| Urgency \| Active failure, deadline, security risk, or release gate \| Needed for the active milestone \| No current pressure \|
26	\| Leverage \| Fits one focused session and prevents rework or creates a reusable result \| Bounded work with moderate payoff \| Unscoped, multi-day, or low-payoff work \|
27
28	5. Break ties by preferring a required prerequisite, then current-evidence
29	verification, then the more reversible action. If the top two remain within
30	one point and target ambiguity would change the answer, run at most three
31	additional read-only checks. If still tied, show both choices and state the
32	single fact that decides between them.
33	6. Turn the recommendation into a 2-4 sentence quick prompt. Keep the scoring
34	and full operator contract internal unless the user asks to compare choices,
35	`expand prompt`, or `make it granular`.
36
37	## Routing Rules
38
39	- If a repo or task already has its own plan, progress doc, issue tracker, or
40	project board, do not create a second one. Treat its recorded next step as
41	one candidate, verify it against current source, and recommend the winner —
42	don't replace the existing tracker.
43	- If the user needs options explored before a commitment can be made, say so
44	and offer to brainstorm instead of forcing a single recommendation.
45	- If missing evidence is the real blocker, make the smallest read-only check
46	the recommended action and generate a prompt for that check.
47
48	## Prompt Levels
49
50	### Quick Prompt (Default)
51
52	Write 2-4 sentences that another capable agent can paste and run:
53
54	1. Name the selected skill or lane, exact target, and one concrete outcome.
55	2. Include the single current fact, scope boundary, or preservation rule most
56	likely to change execution.
57	3. Name the observable done signal and its verification.
58	4. Add a stop condition only when a material blocker is plausible.
59
60	Do not dump field labels, the score breakdown, a long evidence list, or numbered
61	implementation steps by default. After the prompt, add exactly:
62
63	```text
64	Say "expand prompt" for the full operator version or "make it granular" for exact steps and commands.
65	```
66
67	### Expanded Prompt (On Request)
68
69	When the user says `expand prompt`, render the full self-contained prompt below.
70	Include only applicable fields and never leave placeholders such as `TBD` or
71	`<path>`.
72
73	```text
74	Use $<skill> to complete this task.
75
76	Target: <exact repo, worktree, route, URL, document, account, or surface>
77	Objective: <one concrete outcome>
78	Current verified state:
79	- <fresh fact and its source>
80	Authorized scope: <read-only, local edits, live mutation, or other exact boundary>
81	Preserve: <dirty work, copy, data, identities, or other invariants>
82	Source-truth order: <current sources in precedence order>
83
84	Required work:
85	1. <smallest complete sequence>
86
87	Done signal: <observable proof>
88	Verification:
89	- <command, readback, screenshot, URL, test, or response>
90	Stop and report if: <material ambiguity, unsafe mutati

People who liked this also viewed

6 picks

Type

suede-recommend-next-action

44stars8forks

View on GitHub

How to install suede-recommend-next-action?

jasoncolapietro/suede-creator-skills/suede-recommend-next-action

$npx -y skills add JasonColapietro/suede-creator-skills --skill suede-recommend-next-action

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Suede Recommend Next Action
2
3	Recommend one action and package it as a short runnable prompt. Inspect
4	current state read-only; do not execute the recommended action unless the
5	user separately authorizes execution. Keep the full operator contract hidden
6	until the user asks to expand it.
7
8	## Recommendation Workflow
9
10	1. Resolve the target and the user's actual done outcome from the current
11	request, conversation, handoff, plan, repo, or live surface.
12	2. Check only the evidence needed to distinguish the next move. Prefer, in
13	order: current terminal/repo/live state, current source documents, current
14	plans or handoffs, then older memory.
15	3. Generate 2-4 candidate actions internally. Exclude work already verified as
16	complete, adjacent cleanup, and actions outside the user's authorized scope.
17	4. Score each candidate from 0-2 on every criterion below. Recommend the
18	highest total.
19
20	\| Criterion \| 2 points \| 1 point \| 0 points \|
21	\|---\|---\|---\|---\|
22	\| Goal alignment \| Directly produces the user's done signal \| Required prerequisite \| Merely adjacent \|
23	\| Unblocking \| Unlocks a core path or at least two downstream steps \| Unlocks one step \| Unlocks nothing known \|
24	\| Evidence \| Confirmed by current source \| Confirmable with one read-only check \| Depends on an assumption \|
25	\| Urgency \| Active failure, deadline, security risk, or release gate \| Needed for the active milestone \| No current pressure \|
26	\| Leverage \| Fits one focused session and prevents rework or creates a reusable result \| Bounded work with moderate payoff \| Unscoped, multi-day, or low-payoff work \|
27
28	5. Break ties by preferring a required prerequisite, then current-evidence
29	verification, then the more reversible action. If the top two remain within
30	one point and target ambiguity would change the answer, run at most three
31	additional read-only checks. If still tied, show both choices and state the
32	single fact that decides between them.
33	6. Turn the recommendation into a 2-4 sentence quick prompt. Keep the scoring
34	and full operator contract internal unless the user asks to compare choices,
35	`expand prompt`, or `make it granular`.
36
37	## Routing Rules
38
39	- If a repo or task already has its own plan, progress doc, issue tracker, or
40	project board, do not create a second one. Treat its recorded next step as
41	one candidate, verify it against current source, and recommend the winner —
42	don't replace the existing tracker.
43	- If the user needs options explored before a commitment can be made, say so
44	and offer to brainstorm instead of forcing a single recommendation.
45	- If missing evidence is the real blocker, make the smallest read-only check
46	the recommended action and generate a prompt for that check.
47
48	## Prompt Levels
49
50	### Quick Prompt (Default)
51
52	Write 2-4 sentences that another capable agent can paste and run:
53
54	1. Name the selected skill or lane, exact target, and one concrete outcome.
55	2. Include the single current fact, scope boundary, or preservation rule most
56	likely to change execution.
57	3. Name the observable done signal and its verification.
58	4. Add a stop condition only when a material blocker is plausible.
59
60	Do not dump field labels, the score breakdown, a long evidence list, or numbered
61	implementation steps by default. After the prompt, add exactly:
62
63	```text
64	Say "expand prompt" for the full operator version or "make it granular" for exact steps and commands.
65	```
66
67	### Expanded Prompt (On Request)
68
69	When the user says `expand prompt`, render the full self-contained prompt below.
70	Include only applicable fields and never leave placeholders such as `TBD` or
71	`<path>`.
72
73	```text
74	Use $<skill> to complete this task.
75
76	Target: <exact repo, worktree, route, URL, document, account, or surface>
77	Objective: <one concrete outcome>
78	Current verified state:
79	- <fresh fact and its source>
80	Authorized scope: <read-only, local edits, live mutation, or other exact boundary>
81	Preserve: <dirty work, copy, data, identities, or other invariants>
82	Source-truth order: <current sources in precedence order>
83
84	Required work:
85	1. <smallest complete sequence>
86
87	Done signal: <observable proof>
88	Verification:
89	- <command, readback, screenshot, URL, test, or response>
90	Stop and report if: <material ambiguity, unsafe mutati

People who liked this also viewed

6 picks

Type